阅读说明：本技术 工作预测模型的建立和工作推荐方法、装置、设备及介质 (Method, device, equipment and medium for establishing work prediction model and recommending work ) 是由 王超 祝恒书 马超 张敬帅 于 2020-05-28 设计创作，主要内容包括：本申请公开了一种工作预测模型的建立和工作推荐方法、装置、设备及介质,涉及人工智能技术领域。其中,工作预测模型的建立方法的具体实现方案为：根据样本用户简历信息,确定样本用户简历特征表示和样本用户跳槽的目标工作；将样本用户简历特征表示输入原始模型中的时序神经网络,得到用户工作经历表示,将用户工作经历表示输入原始模型中的协同神经网络,得到用户工作特征表示,以及将候选工作的特征表示和用户工作特征表示输入原始模型中的预测神经网络,预测样本用户跳槽到候选工作的概率；根据样本用户跳槽的目标工作和预测结果,对原始模型进行训练,得到工作预测模型。以提高工作推荐的精准性。(The application discloses a method, a device, equipment and a medium for establishing a work prediction model and recommending work, and relates to the technical field of artificial intelligence. The specific implementation scheme of the method for establishing the work prediction model is as follows: determining sample user resume feature representation and target work of sample user slot jumping according to the sample user resume information; inputting the sample user resume feature representation into a time sequence neural network in an original model to obtain a user working experience representation, inputting the user working experience representation into a collaborative neural network in the original model to obtain a user working feature representation, inputting the feature representation of candidate work and the user working feature representation into a prediction neural network in the original model, and predicting the probability of the sample user jumping to the candidate work; and training the original model according to the target work and prediction results of the sample user slot jumping to obtain a work prediction model. So as to improve the accuracy of work recommendation.)

1. A method for establishing a working prediction model comprises the following steps:

determining sample user resume feature representation and target work of sample user slot jumping according to the sample user resume information;

inputting the sample user resume feature representation into a time sequence neural network in an original model to obtain a user working experience representation, inputting the user working experience representation into a collaborative neural network in the original model to obtain a user working feature representation, and inputting the feature representation of candidate work and the user working feature representation into a prediction neural network in the original model to predict the probability of the sample user jumping to the candidate work;

and training the original model according to the target work and prediction results of the sample user slot jumping to obtain a work prediction model.

2. The method of claim 1, wherein a sub-sequential network of the sequential neural networks has a one-to-one correspondence with a sub-aware network of the cooperative neural networks;

the first output end of the sub-time sequence network is connected with the next sub-time sequence network;

and the second output end of the sub time sequence network is connected with the corresponding sub perception network.

3. The method of claim 2, wherein the sub-sensing network comprises at least one of a first sub-sensing unit, a second sub-sensing unit, and a third sub-sensing unit; the input ends of the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are all connected with the second output end of the corresponding sub-timing network, and the output ends of the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are all connected with the input end of the prediction neural network;

the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are respectively used for outputting user time characteristic representation, user company characteristic representation and user post characteristic representation in the user work characteristic representation.

4. The method of claim 2, wherein different sub-aware networks in the cooperative neural network share model parameters.

5. The method of claim 2, wherein the sub-timing network is a long short term memory network (LSTM); the sub-perception network is a multi-layer perception machine MLP.

6. The method of any of claims 1-5, wherein determining a sample user resume feature representation from sample user resume information comprises:

determining sample user attribute information, at least one piece of work attribute information of the sample user and the work duration of the work according to the sample user resume information;

determining static attribute feature representation of the sample resume according to the sample user attribute information;

determining dynamic attribute feature representation of the sample resume according to the working attribute information of at least one job of the sample user;

encoding the working duration of at least one part of work of the sample user, and determining a sample time encoding representation;

and taking the sample resume static attribute feature representation, the sample resume dynamic attribute feature representation and the sample time coding representation as the sample user resume feature representation.

7. The method of claim 6, wherein inputting the sample user resume feature representation into a temporal neural network in an original model comprises:

obtaining an initial transmission parameter according to the static attribute feature representation of the sample resume;

taking the initial transfer parameter as a transfer input of a first sub-time sequence network of a time sequence neural network in the original model;

and sequentially taking the dynamic attribute feature representation and the time coding representation of the sample resume of each job and the static attribute feature representation of the sample resume as the parameter input of a sub time sequence network in the time sequence neural network.

8. The method of claim 1, wherein inputting the representation of the characteristics of the candidate job and the representation of the user job characteristics into a predictive neural network in the original model, predicting a probability of the sample user jumping to the candidate job, comprises:

inputting the feature representation of the candidate work, the slot jumping time and the user work feature representation into a prediction neural network in the original model, and predicting the probability of the sample user jumping to the candidate work.

9. The method of claim 8, wherein the predictive neural network comprises a slot jumping company predictive subnetwork and/or a slot jumping position predictive subnetwork;

correspondingly, inputting the feature representation of the candidate work, the slot jump time and the user work feature representation into a prediction neural network in the original model, and predicting the probability of the sample user jumping to the candidate work comprises:

inputting a feature representation of a candidate company in the feature representation of the candidate work, a slot jumping time and a user company feature representation in the user work feature representation into the slot jumping company prediction sub-network, and predicting the probability of the sample user jumping to the candidate company; and/or the presence of a gas in the gas,

inputting the feature representation of the candidate position in the feature representation of the candidate work, the slot jumping time and the user position feature representation in the user work feature representation into the slot jumping position prediction sub-network, and predicting the probability of the sample user jumping to the candidate position.

10. The method of any one of claims 8-9, wherein the predictive neural network further comprises: a slot hop time prediction subnetwork;

correspondingly, inputting the feature representation of the candidate work, the slot jumping time and the user work feature representation into the prediction neural network in the original model, and before predicting the probability of the sample user jumping to the candidate work, the method further comprises:

and inputting the user time characteristic representation in the user working characteristic representation into the slot-skipping time prediction sub-network, and predicting to obtain the slot-skipping time.

11. A work recommendation method implemented using a work prediction model established by the method of any one of claims 1-10, the method comprising:

determining the resume characteristic representation of the target user according to the resume information of the target user;

inputting the target user resume feature representation into the work prediction model to obtain the probability of the target user jumping to the candidate work;

and recommending work for the target user according to the probability of the candidate work for the user to jump the slot.

12. The method of claim 11, wherein determining a target user resume feature representation from target user resume information comprises:

determining attribute information of the target user, working attribute information of at least one part of work of the target user and working duration of the part of work according to the resume information of the target user;

determining static attribute feature representation of the target resume according to the attribute information of the target user;

determining dynamic attribute feature representation of the target resume according to the working attribute information of at least one part of work of the target user;

coding at least one part of working time of the target user, and determining target time coding expression;

and taking the target resume static attribute feature representation, the target resume dynamic attribute feature representation and the target time coding representation as the target user resume feature representation.

13. The method of claim 12, wherein inputting the user resume feature representation into the work prediction model comprises:

obtaining an initial transfer parameter according to the static attribute feature representation of the target resume;

taking the initial transfer parameter as a transfer input of a first sub-time sequence network of a time sequence neural network in the work prediction model;

and sequentially taking the dynamic attribute feature representation and the target time coding representation of the target resume of each job and the static attribute feature representation of the target resume as the parameter input of a sub time sequence network in the time sequence neural network.

14. An apparatus for creating a working prediction model, comprising:

the sample preprocessing module is used for determining sample user resume feature representation and target work of sample user slot jumping according to the sample user resume information;

the model training module is used for inputting the sample user resume feature representation into a time sequence neural network in an original model to obtain a user work experience representation, inputting the user work experience representation into a collaborative neural network in the original model to obtain a user work feature representation, inputting the feature representation of candidate work and the user work feature representation into a prediction neural network in the original model, and predicting the probability of the sample user jumping to the candidate work; and training the original model according to the target work and prediction results of the sample user slot jumping to obtain a work prediction model.

15. The apparatus of claim 14, wherein a sub-sequential network of the sequential neural networks has a one-to-one correspondence with a sub-aware network of the cooperative neural networks;

the first output end of the sub-time sequence network is connected with the next sub-time sequence network;

and the second output end of the sub time sequence network is connected with the corresponding sub perception network.

16. The apparatus according to claim 15, wherein the sub-sensing network comprises at least one of a first sub-sensing unit, a second sub-sensing unit, and a third sub-sensing unit; the input ends of the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are all connected with the second output end of the corresponding sub-timing network, and the output ends of the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are all connected with the input end of the prediction neural network;

the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are respectively used for outputting user time characteristic representation, user company characteristic representation and user post characteristic representation in the user work characteristic representation.

17. The apparatus of claim 15, wherein different sub-aware networks in the cooperative neural network share model parameters.

18. The apparatus of any of claims 14-17, wherein the sample pre-processing module comprises:

the information determining unit is used for determining the attribute information of the sample user, the working attribute information of at least one part of work of the sample user and the working duration of the part of work according to the resume information of the sample user;

the static characteristic determining unit is used for determining the static attribute characteristic representation of the sample resume according to the sample user attribute information;

the dynamic characteristic determining unit is used for determining dynamic attribute characteristic representation of the sample resume according to the working attribute information of at least one job of the sample user;

the time coding unit is used for coding the working time of at least one part of work of the sample user and determining the sample time coding expression;

and the characteristic integration unit is used for taking the sample resume static attribute characteristic representation, the sample resume dynamic attribute characteristic representation and the sample time coding representation as the sample user resume characteristic representation.

19. The apparatus of claim 18, wherein the model training module comprises a first data input unit for inputting the sample user resume feature representations into a temporal neural network in a raw model; the first data input unit is specifically configured to:

obtaining an initial transmission parameter according to the static attribute feature representation of the sample resume;

taking the initial transfer parameter as a transfer input of a first sub-time sequence network of a time sequence neural network in the original model;

and sequentially taking the dynamic attribute feature representation and the time coding representation of the sample resume of each job and the static attribute feature representation of the sample resume as the parameter input of a sub time sequence network in the time sequence neural network.

20. The apparatus of claim 14, wherein the model training module comprises a second data input unit for inputting a feature representation of a candidate job and the user job feature representation into a predictive neural network in the original model, predicting a probability of the sample user jumping to the candidate job; the second data input unit is specifically configured to:

inputting the feature representation of the candidate work, the slot jumping time and the user work feature representation into a prediction neural network in the original model, and predicting the probability of the sample user jumping to the candidate work.

21. The apparatus of claim 20, wherein the predictive neural network comprises a slot jumping company predictive subnetwork and/or a slot jumping position predictive subnetwork;

correspondingly, the second data input unit is specifically configured to:

inputting a feature representation of a candidate company in the feature representation of the candidate work, a slot jumping time and a user company feature representation in the user work feature representation into the slot jumping company prediction sub-network, and predicting the probability of the sample user jumping to the candidate company; and/or the presence of a gas in the gas,

inputting the feature representation of the candidate position in the feature representation of the candidate work, the slot jumping time and the user position feature representation in the user work feature representation into the slot jumping position prediction sub-network, and predicting the probability of the sample user jumping to the candidate position.

22. The apparatus of any one of claims 20-21, wherein the predictive neural network further comprises: a slot hop time prediction subnetwork;

correspondingly, the model training module further comprises:

and the third data input unit is used for inputting the user time characteristic representation in the user work characteristic representation into the groove jumping time prediction sub-network to predict the groove jumping time.

23. A work recommendation apparatus implemented using a work prediction model established by the method of any one of claims 1-10, the apparatus comprising:

the user resume processing module is used for determining the resume characteristic representation of the target user according to the resume information of the target user;

the work prediction module is used for inputting the target user resume feature representation into the work prediction model to obtain the probability of the target user jumping to the candidate work;

and the work recommending module is used for recommending work for the target user according to the probability of the candidate work for the user to jump the slot.

24. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of building a working prediction model according to any of claims 1-10; or to perform the work recommendation method of any of claims 11-13.

25. A non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method of building a working prediction model of any one of claims 1-10; or to perform the work recommendation method of any of claims 11-13.

Technical Field

The application relates to the technical field of computers, in particular to a work prediction model establishing method and a work recommendation method by using an artificial intelligence technology.

Background

Under the fast-paced environment of the current society, the work mobility is greatly increased compared with the past, more and more work opportunities are presented to people, and job seekers are often trapped in mass job information on the Internet and are difficult to select. Therefore, a Job recommendation System (Job recommendation System) for recommending appropriate jobs to the Job seekers is gradually emerging. At present, the existing job recommendation system generally performs preliminary job screening for job seekers based on conditions such as geographical locations, salary levels, working hours and the like selected by the job seekers, and then sorts the screened jobs according to the working heat of the current time period and recommends the sorted jobs to the job seekers. The existing work recommendation system cannot recommend based on the personalized requirements of job seekers, and the recommendation result is not accurate enough.

Disclosure of Invention

A method, an apparatus, a device and a medium for establishing a work prediction model and recommending work are provided.

According to an aspect of the present disclosure, there is provided a method of establishing a work prediction model, the method including:

determining sample user resume feature representation and target work of sample user slot jumping according to the sample user resume information;

inputting the sample user resume feature representation into a time sequence neural network in an original model to obtain a user working experience representation, inputting the user working experience representation into a collaborative neural network in the original model to obtain a user working feature representation, inputting the feature representation of candidate work and the user working feature representation into a prediction neural network in the original model, and predicting the probability of the sample user jumping to the candidate work;

and training the original model according to the target work and prediction results of the sample user slot jumping to obtain a work prediction model.

According to a second aspect, there is provided a work recommendation method implemented using a work prediction model established by the method of any embodiment of the present application, the method comprising:

determining the resume characteristic representation of the target user according to the resume information of the target user;

inputting the resume feature representation of the target user into a work prediction model to obtain the probability of the target user jumping to the candidate work;

and recommending work for the target user according to the probability of the candidate work of the user jumping the slot.

According to a third aspect, there is provided an apparatus for creating a working prediction model, the apparatus comprising:

the sample preprocessing module is used for determining sample user resume feature representation and target work of sample user slot jumping according to the sample user resume information;

the model training module is used for inputting the sample user resume feature representation into a time sequence neural network in the original model to obtain a user work experience representation, inputting the user work experience representation into a collaborative neural network in the original model to obtain a user work feature representation, inputting the feature representation of candidate work and the user work feature representation into a prediction neural network in the original model, and predicting the probability of the sample user jumping to the candidate work; and training the original model according to the target work and prediction results of the sample user slot jumping to obtain a work prediction model.

According to a fourth aspect, there is provided a work recommendation apparatus implemented using a work prediction model established by the method of any embodiment of the present application, the apparatus comprising:

the user resume processing module is used for determining the resume characteristic representation of the target user according to the resume information of the target user;

the work prediction module is used for inputting the resume feature representation of the target user into the work prediction model to obtain the probability of the target user jumping to the candidate work;

and the work recommending module is used for recommending work for the target user according to the probability of the candidate work of the user jumping the slot.

According to a fifth aspect, there is provided an electronic device comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a method of creating a work prediction model or a method of work recommendation according to any of the embodiments of the present application.

According to a sixth aspect, a non-transitory computer readable storage medium having computer instructions stored thereon is provided. The computer instructions are used for causing a computer to execute the work prediction model establishing method or the work recommendation method of any embodiment of the application.

According to the technology of the embodiment of the application, the problem that the conventional work recommendation system cannot perform work recommendation based on the individualized requirements of job seekers is solved, and the accuracy of work recommendation is improved.

It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present disclosure, nor do they limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.

Drawings

The drawings are included to provide a better understanding of the present solution and are not intended to limit the present application. Wherein:

FIG. 1A is a flow chart of a method for establishing a work prediction model according to an embodiment of the present application;

FIG. 1B is a schematic diagram of a structure of an original model provided according to an embodiment of the present application;

FIG. 2 is a schematic structural diagram of another original model provided according to an embodiment of the present application;

FIG. 3A is a flow chart of another method for building a work prediction model according to an embodiment of the present application;

FIG. 3B is a schematic diagram illustrating an operation principle of a time-series neural network of an original model according to an embodiment of the present application;

FIG. 4A is a flow chart of another method for building a work prediction model according to an embodiment of the present application;

FIG. 4B is a schematic diagram illustrating the operation of a predictive neural network of an original model according to an embodiment of the present application;

FIG. 5A is a flowchart of a work recommendation method provided in accordance with an embodiment of the present application;

FIG. 5B is a schematic diagram of a structure of a working prediction model provided according to an embodiment of the present application;

FIG. 6 is a schematic structural diagram of an apparatus for building a work prediction model according to an embodiment of the present application;

FIG. 7 is a schematic structural diagram of a work recommendation device according to an embodiment of the present application;

fig. 8 is a block diagram of an electronic device for implementing a work prediction model building method or a work recommendation method according to an embodiment of the present application.

Detailed Description

The following description of the exemplary embodiments of the present application, taken in conjunction with the accompanying drawings, includes various details of the embodiments of the application for the understanding of the same, which are to be considered exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present application. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

FIG. 1A is a flow chart of a method for establishing a work prediction model according to an embodiment of the present application; FIG. 1B is a schematic diagram of a structure of an original model provided according to an embodiment of the present application; the embodiment is suitable for the condition of constructing and training the neural network model capable of executing the work prediction task. The embodiment may be performed by an apparatus for establishing a working prediction model configured in an electronic device, and the apparatus may be implemented by software and/or hardware. As shown in fig. 1A-1B, the method includes:

and S101, determining sample user resume feature representation and target work of sample user slot jumping according to the sample user resume information.

The sample user resume information may be text information contained in the user resume as the sample, and optionally, the sample user resume information may include personal attribute information and work experience information in the user resume. For example, the personal attribute information may include gender, age, self-assessment, etc. of the user, and the work history information may include the work duration and work attribute information for each work of the user, such as the work attribute information may in turn include the size, type, address, profile, etc. of the work unit. The user resume feature representation may be a user resume feature characterized in a form of numbers or letters after encoding the sample user resume information of the text class. The user resume feature representation can be represented in the form of a vector or a matrix, and the like. The target work of the sample user for jumping the slot may be determined according to the work experience of the sample user resume, and specifically, the next work in the work experience may be sequentially used as the target work of the sample user for jumping the slot corresponding to the previous work. For example, if there are three jobs in the sample user resume, the second job may be a target job for the sample user to skip the slot corresponding to the first job; the third job can be used as a target job of slot skipping of the sample user corresponding to the second job. The target work of the sample user jumping the slot may include, but is not limited to: the time the sample user jumps to the target job, the company and post the target job, etc.

Optionally, in this embodiment of the application, the sample user resume information of the text type may be encoded into a sample user resume feature representation of a number or letter type according to a preset encoding rule, and a second job to a last job in the job experience of the sample user resume are respectively used as target jobs for the sample user to jump to the slot corresponding to the previous job. Optionally, the preset encoding rule may be that numerical value information in the sample user resume information, such as the age and the working age of the sample user or the registration duration of each company in which the sample user has worked, is directly used as the sample resume feature representation. The character information in the user resume information is converted into a standard numerical value class vector or matrix sample user resume characteristic representation by adopting a word segmentation processing technology and a word vector coding technology (such as a word2vec technology).

Optionally, since data in the collected sample user resume information is usually relatively cluttered, in the embodiment of the present application, the sample user resume information may be preprocessed in advance, and after abnormal data (such as abnormal symbols) therein is deleted, the operation of determining sample user resume feature representation and sample user slot skipping target work in the step is performed on the remaining sample user resume information, so as to improve the accuracy of the determined sample user resume feature representation and target work.

S102, inputting the sample user resume feature representation into a time sequence neural network in an original model to obtain a user work experience representation, inputting the user work experience representation into a collaborative neural network in the original model to obtain a user work feature representation, inputting the feature representation of candidate work and the user work feature representation into a prediction neural network in the original model, and predicting the probability of the sample user jumping to the candidate work.

The so-called original model may be a constructed, but untrained, working prediction model. Optionally, as shown in fig. 1B, the original model 1 in the embodiment of the present application includes three parts, namely, a time-series neural network 10, a cooperative neural network 11, and a prediction neural network 12. The input of the original model 1 is the input of the time sequence neural network 10, the output of the time sequence neural network 10 is connected with the input of the cooperative neural network 11, the output of the cooperative neural network 11 is connected with the input of the prediction neural network 12, and the output of the prediction neural network 12 is the output of the original model 1.

Optionally, in this embodiment of the application, the user work experience expression may be a hidden variable feature corresponding to each work of the sample user, which is extracted from the sample user resume feature; the user work feature representation may be the hidden variable feature of the user further extracted according to the hidden variable feature corresponding to each work. Among the massive sample users, the user work characteristic representations of similar sample users are also similar. The feature representation of the candidate job may be an implicit variable feature of the candidate job determined for each candidate job, where, in the massive candidate jobs, the feature representations of the candidate jobs of similar candidate jobs (such as similar companies and/or similar posts) are also similar.

Optionally, in this embodiment of the present application, the sample user resume features determined in S101 may be input into the time-series neural network 10 of the original model 1, and the time-series neural network 10 may analyze the input sample user resume feature representations and output user work experience representations, where the user experience representations may include sub-work experience representations corresponding to each work included in the work experience of the sample user resume. The user work experience representation output by the time sequence neural network 10 is further input into the collaborative neural network 11, the collaborative neural network 11 performs collaborative perception on the sub-work experience representation corresponding to each work, outputs the user work characteristic representation corresponding to each work, and further inputs the user work characteristic representation into the prediction neural network 12, the prediction neural network 12 combines the user work characteristic representation corresponding to each work and the characteristic representation of each candidate work for analysis, and outputs the probability of jumping from the work to each candidate work.

Optionally, the probability of the sample user jumping to the candidate job predicted by the embodiment of the present application may include, but is not limited to: the time from the sample user to the candidate job, the candidate company of the slot, the candidate post of the slot, etc.

S103, training the original model according to the target work and prediction result of the sample user jumping the slot to obtain a work prediction model.

In S102, the prediction result is the probability that the prediction neural network in the original model predicts the job in the sample user resume and jumps from the job to each candidate job.

Optionally, in this step, for each sample user, the probability of each predicted job of the sample user from the job slot to each candidate job, except the last job, and the target job of the sample user from the job slot (i.e., the next job of the job) are used as a set of training data. And continuously training the original model by adopting a gradient descent method based on a plurality of groups of training data of each sample user, and continuously updating and optimizing network parameters of a time sequence neural network, a cooperative neural network and a prediction neural network in the original model until the model converges to obtain a working prediction model.

Optionally, in the embodiment of the application, after the training of the original model reaches a preset duration or a preset number of times, the test data may be used to perform the prediction precision test on the trained original model, and if the precision of the trained original model meets a preset requirement, the trained original model is the work prediction model.

Optionally, the trained work prediction model may be embedded into a work recommendation system, and used for online predicting probability values of target users (such as job seekers) jumping to each candidate work. Specifically, the resume can be uploaded to a work recommendation system by a target user, the work recommendation system can determine resume feature representation of the target user according to resume features of the target user, then the resume feature representation of the target user is input into a work prediction model, so that the probability of each candidate work of the target user from the current work skip of the target user predicted by the work prediction model is obtained, and one or more candidate works with higher skip probability are selected as the skip work to be recommended to the user.

According to the technical scheme, an original model comprising a time sequence neural network, a cooperative perception network and a prediction neural network is constructed, user resume characteristic representation determined based on sample user resume information is input to the time sequence neural network of the original model, user working experience representation is input to the cooperative neural network, user working characteristic representation is input to the prediction neural network, the probability that a sample user jumps to candidate work is predicted, and the original model is trained by combining target work of actual groove jumping of the sample user to obtain a working prediction model. The working prediction model is trained on resume information of a large number of different sample users, so that when the working prediction model is used for predicting a certain user on a groove jumping working line in the subsequent process, the probability from the groove jumping of the target user to each candidate work can be determined by cooperating with the selection of other users similar to the working characteristic representation of the target user in the groove jumping process. The accuracy of the prediction result is greatly improved while work recommendation based on the personalized requirements of the user can be realized.

Fig. 2 is a schematic structural diagram of another original model provided according to an embodiment of the present application. On the basis of the above embodiments, the present embodiment further optimizes the structure of the constructed original model, and provides a description of a specific situation of the internal structures of the time-series neural network and the cooperative neural network of the original model. Specifically, the time sequence neural network of the original model comprises a plurality of sub time sequence networks, the cooperative neural network in the original model comprises a plurality of sub perception networks, and the sub time sequence networks in the time sequence neural network correspond to the sub perception networks in the cooperative neural network one to one; the first output end of the sub-time sequence network is connected with the next sub-time sequence network; and the second output end of the sub time sequence network is connected with the corresponding sub perception network. Optionally, the sub-timing network is a long-short term memory network LSTM; the sub-perception network is a multi-layer perception machine MLP. Illustratively, as shown in FIG. 2, the time-series neural network 10 of the original model 1 includes T sub-time-series networks 101, namely LSTM-1 to LSTM-T. The cooperative neural network 11 in the original model 1 includes T sub-perception networks 111, namely MLP-1 to MLP-T. And LSTM-1 corresponds to MLP-1, LSTM-2 corresponds to MLP-2, and … LSTM-T corresponds to MLP-T. For each LSTM, a first output end A of the LSTM is connected with a transmission input end C of the next LSTM; the second output terminal B of the LSTM is connected to the input terminal of its corresponding MLP. Each LSTM has, in addition to the delivery input C, a parameter input D for inputting a sample user profile representation corresponding to each job.

The time sequence modeling method is introduced to construct an original model comprising a plurality of sub time sequence networks, so that the whole work experience link of a user can be completely analyzed, and the work change process information of the user is mined. Meanwhile, each sub time sequence network corresponds to one sub perception network, the working characteristic representation of the user can be accurately extracted, and because the working characteristics of the users of similar users are similar, the embodiment of the application can provide suggestions for the professional development of the users in an auxiliary mode by utilizing the selection information of the similar users on different positions, and further more accurate working prediction results are obtained.

Further, in this embodiment of the application, the cooperative neural network of the original model includes a plurality of sub-sensing networks, and each sub-sensing network further includes at least one of a first sub-sensing unit, a second sub-sensing unit, and a third sub-sensing unit; the input ends of the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are all connected with the second output end of the corresponding sub-timing network, and the output ends of the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are all connected with the input end of the prediction neural network; the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are respectively used for outputting user time characteristic representation, user company characteristic representation and user post characteristic representation in the user work characteristic representation. Optionally, for each sub-sensing unit, it may be a set of multi-layer sensing machines MLP.

Exemplarily, as shown in fig. 2, a sub-sensing network MLP-1 is taken as an example, and includes three groups of sub-sensing units, namely, a first sub-sensing unit corresponding to a diagonal square frame, a second sub-sensing unit corresponding to a vertical square frame, and a third sub-sensing unit corresponding to a horizontal square frame. Optionally, the three groups of sub-sensing units may be formed by three groups of multi-layer sensors, each group of multi-layer sensors including an input layer, at least one hidden layer and an output layer. The input layers of the sub-sensing units are the same and are connected with the second output end B of the sub-time sequence network, and the output layer of each sub-sensing unit is connected with the input end of the prediction neural network. Wherein, the first sub-sensing unit of the sub-sensing network MLP-1 is used for outputting the user time characteristic representation x in the user work characteristic representation through multi-layer sensing_i1(i.e., the user time signature corresponding to the 1 st job of the ith user); the second sub-perception unit is used for outputting a user company feature representation u in the user work feature representation through multi-layer perception_i1(i.e., the user company characterization representation corresponding to the 1 st job of the ith user); the third sub-sensing unit 1113 is configured to output the user position feature representation w in the user work feature representation through multi-layer sensing_i1(i.e. the user station characteristic table corresponding to the 1 st job of the ith userShown).

It should be noted that, in this embodiment of the application, each sub-sensing network may include at least one of the three sub-sensing units, which is not limited in this embodiment. The user work characteristic representations output by different sub-sensing units are used for predicting different groove jumping work categories, for example, the user time characteristic representation output by a first sub-sensing unit is used for predicting groove jumping time, the user company characteristic representation output by a second sub-sensing unit is used for predicting groove jumping companies, and the user position characteristic representation output by a third sub-sensing unit is used for measuring groove jumping positions. The method and the device have the advantages that the multidimensional user working characteristic representation can be extracted through the perception network, and therefore the more accurate groove jumping work for the user prediction from the multiple dimensions is achieved.

Optionally, for each sub-sensing network or each sub-sensing unit, the specific working principle may be as follows: the user work experience representation of the Tth work of the user i transmitted by the sub time sequence network is received by each sub sensing network or sub sensing unitThereafter, it can be determined by its internal multi-layer perceptron, i.e. at least one hidden layerThe corresponding user job feature representation. Specifically, each hidden layer may have two parameters, namely weight coefficients and a bias vector. For the t hidden layer, assume its weight coefficient is W_tAnd the bias vector is b_tThen, can be according to formula g_t＝f_t(W_tg_t-1+b_t)，t∈[2,n-1]Determining a perception result g of a t-th hidden layer_t. Wherein n is the number of network layers of the sub-sensing network or the sub-sensing unit, g_t-1As a result of the perception of the t-1 th hidden layer, f_t() Is the activation function of the t-th hidden layer. Optionally, for different hidden layers, the activation function may be selected according to requirements, for example, a Sigmoid activation function is adopted for the first n-1 hidden layers, and the pairThe last hidden layer takes the tanh activation function.

Optionally, in this embodiment of the present application, different sub-perception networks in the cooperative neural network share a model parameter. The model parameters may include, but are not limited to, weighting coefficients and bias vectors of the network. Specifically, the cooperative neural network comprises a plurality of sub-perception networks, and all the sub-perception networks share the weight coefficient and the offset vector of the same multi-layer perception mechanism. Optionally, when each sub-sensing network includes a plurality of sub-sensing units, the sub-sensing units of the same type share the same weight coefficient and bias vector. The advantage of such an arrangement is embodied in two aspects, in the first aspect, since the process of obtaining the user working characteristic representation by the collaborative neural network analysis is independent from time, the similar user working characteristic representation represents a similar user, and even if the two are not in the same time period, the corresponding users are the same, so that the sub-perception networks with the same model parameters need to be used for processing the user working experience representations output by the different sub-timing networks. On the other hand, different sub-perception networks share model parameters, so that parameters needing to be learned of the cooperative neural network can be greatly reduced, the robustness of the model is enhanced, and the risk of overfitting is reduced.

Fig. 3A is a flowchart of another method for establishing a work prediction model according to an embodiment of the present application. Fig. 3B is a schematic diagram illustrating an operation principle of a time-series neural network of an original model according to an embodiment of the present application. Based on the above embodiments, the present embodiment performs further optimization, and gives a specific description on how to determine the sample user resume feature representation according to the sample user resume information and input the sample user resume feature representation into the original model. As shown in fig. 3A-3B, the method includes:

s301, according to the sample user resume information, determining sample user attribute information, at least one piece of work attribute information of the sample user and the work duration of the work.

The sample user attribute information may be related information characterizing the static attributes of the sample user, and may include, for example, but not limited to, the age, sex, work age, personal evaluation, and the like of the sample user. The job attribute information may be information related to each job in the sample user job experience, and may include, for example, but not limited to, the size, type, job position, annual outflow and inflow rates of the company employees, and company introduction text, among others.

Optionally, in the embodiment of the application, the sample user resume information may be classified, and information representing the age, sex, working age, personal evaluation and the like of the user static attribute information is screened out from the sample user resume information as the sample user attribute information. Because the work experience in the sample resume information usually includes at least one job, this embodiment may be to screen information, such as company scale, type, employment post, annual outflow and inflow rate of company employees, and company introduction text, which characterize each job from the sample user resume as work attribute information, and screen the work duration of each job of the user.

S302, determining the static attribute feature representation of the sample resume according to the sample user attribute information.

Optionally, because the sample user attribute information includes two kinds of attribute information, namely a text type attribute information and a numerical value type attribute information, the step may be that for the numerical value type sample user attribute information, such as the age, the age and the expected wage of the sample user, the corresponding numerical value is directly represented in the form of a vector or a matrix, and is used as the static attribute feature representation of the sample resume; for sample user attribute information of characters, such as gender, personal evaluation text content and the like, word segmentation is carried out by adopting a word segmentation technology, then word vector coding technology (such as word2vec technology) is adopted, words after word segmentation are coded into numerical values and then expressed in a vector or matrix form, and the numerical values are used as sample resume static attribute feature expression.

Optionally, since abnormal data, such as abnormal symbols, may exist in the sample user attribute information determined in S301, in this step, the sample user attribute information determined in S301 may be cleaned first, and after the abnormal data included in the sample user attribute information is deleted, the operation of determining the static attribute feature representation of the sample resume in this step is performed.

S303, determining dynamic attribute feature representation of the sample resume according to the work attribute information of at least one work of the sample user.

Optionally, for the work attribute information of each job of the sample user, the work attribute information is similar to the sample user attribute information and includes both attribute information of a text type and attribute information of a numerical type, so this step may be similar to S302, and for numerical type information in the work attribute information of each job, such as company scale, annual outflow and inflow rates of company employees, company creation year, and the like, the corresponding numerical value is directly represented in a vector or matrix form as a sample resume dynamic attribute feature representation; for the working attribute information of the characters, such as formula types, addresses, company introduction texts and the like, word segmentation is carried out by adopting a word segmentation technology, and then word vector coding technology (such as word2vec technology) is adopted, the word groups after word segmentation are coded into numerical values and then are expressed in the form of vectors or matrixes to be used as sample resume dynamic attribute feature expression.

Optionally, in the same way, abnormal data, such as abnormal symbols, may also exist in the work attribute information, so the step may adopt a method similar to the above-mentioned S302, and the work attribute information of each work determined in S301 is cleaned first, and after the abnormal data contained therein is deleted, the operation of determining the dynamic attribute characteristic representation of the sample resume in the step is executed.

S304, encoding the working time of at least one part of work of the sample user, and determining the sample time encoding expression.

Optionally, the working duration (e.g., one year, two years, etc.) of each job of the sample user belongs to a one-dimensional discrete time variable, and in order to deeply utilize the deep analysis capability of the neural network, the working duration of each job of the sample user needs to be encoded and converted into a time hidden variable in a high-dimensional continuous space, that is, a sample time encoding representation. It should be noted that, in the process of encoding the working duration in this step, the characteristics of the original time value need to be maintained, for example, the euclidean distances corresponding to the sample time encoding representations after the working duration is encoded for 3 years and 4 years in the high-dimensional space are closer than the euclidean distances corresponding to the sample time encoding representations after the working duration is encoded for 1 year and 6 years in the high-dimensional space.

Optionally, in this embodiment of the application, the working duration may be encoded by using two implementable manners, i.e., a preset algorithm and a model learning method, to obtain a sample time encoding representation. Specifically, the method comprises the following steps:

in the first implementation mode, the working duration is encoded by using a preset algorithm, and when the sample time encoding representation is determined, the one-dimensional discrete working duration is encoded into the high-dimensional continuous sample time encoding representation according to a preset formula, such as the following formulas (1) - (2), by analyzing the time information.

Wherein D is_iTThe working time of the Tth work of the user i;andrespectively represent D_iTThe values are encoded at sample time in dimensions 2j and 2j +1, and r is the size of the dimension represented by the sample time encoding.

For example, assuming that the one-dimensional discrete operating time length is encoded into a 5-dimensional continuous time sample encoding representation, then r is 5, and the 1, 3, 5-dimensional corresponding sample time encoding values can be calculated by using formula (1); and (3) calculating sample time coding values corresponding to 2 and 4 dimensions by adopting a formula (2), and representing the calculated 5-dimensional sample time coding values as coded 5-dimensional continuous sample time coding. The advantage of determining the sample time coding representation by adopting the implementable mode is that more calculation amount is not needed to learn the coding parameters, and the coding process is simple and convenient.

In the second implementation mode, the working time length is encoded by using a model learning method, and when the sample time code expression is determined, the encoding process may be integrated in the whole learning training process of the original model, specifically, a time code expression is established for each working time length and paired with the time code expression, and then in the subsequent training process of the original model, the encoding parameters between the working time length and the paired time code expressions are continuously updated. A benefit of using this implementable embodiment is that the determined sample time encoding represents a time encoding that is more consistent with the real data distribution.

S305, taking the sample resume static attribute feature representation, the sample resume dynamic attribute feature representation and the sample time coding representation as sample user resume feature representation.

Optionally, in this embodiment of the application, the sample resume static attribute feature representation determined in S302, the sample resume dynamic attribute feature representation determined in S303, and the sample time code representation determined in S304 are collectively used as the sample user resume feature representation.

And S306, determining the target work of the sample user for jumping the slot according to the sample user resume information.

S307, inputting the sample user resume feature representation into a time sequence neural network in the original model to obtain a user working experience representation.

Optionally, in this embodiment, the initial transfer parameter may be obtained according to the static attribute feature representation of the sample resume; taking the initial transmission parameters as transmission input of a first sub time sequence network of the time sequence neural network in the original model; and sequentially taking the dynamic attribute feature representation and the time code representation of the sample resume of each job and the static attribute feature representation of the sample resume as the parameter input of a sub-time sequence network in the time sequence neural network. Specifically, as shown in fig. 3B, a sample user resume static attribute feature in the sample user resume feature may be represented by F_isMapping to obtain initial transmission parametersOptionally, F can be_isInput to a data mapping layer, which may map F_isTo carry outObtained after mapping processingThen will beAs the delivery input to the first sub-timing network LSTM-1. Then sequentially representing the dynamic characteristics of the resume of each work sample F_iTThe working time length D of the work_iTIs represented by a sample time code d_iTAnd sample resume static attribute feature representation F_isAdded hidden variablesAs a parameter input to a sub-timing network LSTM. For example, F may be the first job_i1The working time length D of the job_i1Is represented by a sample time code d_i1And sample resume static attribute feature representation F_isAdded hidden variablesAs parameter input of LSTM-1; working the second part of F_i2The working time length D of the job_i2Is represented by a sample time code d_i2And sample resume static attribute feature representation F_isImplicit variable after additionInputting parameters of LSTM-2, and analogizing to work on the T th work and work on the T th work_iTThe working time length D of the job_iTIs represented by a sample time code d_iTAnd sample resume static attribute feature representation F_isAdded hidden variablesAnd the parameter input of the LSTM-T is input into an input end D of the LSTM-T.

Optionally, for each sub-sequential network in the sequential neural networkThe input comprises two parts, one part is the transfer input, namely the transfer parameter of the last sub-sequence network outputThe other part is parameter inputThe output also comprises two parts, one part is the transmission parameter of the current sub-time sequence network outputAnother part is the representation of the user's work experience of the current sub-sequential network outputOptionally, passing the inputComprisesThe related information of (2). When the sub-time sequence network is an LSTM network, the formula can be shownTo determine the output of the current sub-sequential networkAndwherein the content of the first and second substances,used for outputting to the sub-perception network corresponding to the current sub-timing network,used as the transmission parameter of the current sub-time sequence network and output to the next sub-time sequence network.

It should be noted that, in the embodiment of the present application, a process of obtaining an initial delivery parameter according to a sample resume static attribute feature representation, and a process of representing a sample resume dynamic attribute feature representation and a sample time code representation of each job, that is, a process in fig. 3BAnd d_i1-d_iTThe determination process can be executed by a data processing layer in the original model, so long as the sample resume static attribute feature is represented by F_isSample resume dynamic characterization for each job F_iTAnd the working time length D of the work_iTAnd inputting the original model. The initial transfer parameter and the sample time coding representation of each operating time may also be determined before the original model is input, and the initial transfer parameter and the sample time coding representation may be directly input to the original model, which is not limited in this embodiment.

And S308, inputting the user work experience representation into the collaborative neural network in the original model to obtain the user work characteristic representation.

S309, inputting the characteristic representation of the candidate work and the user work characteristic representation into a prediction neural network in the original model, and predicting the probability of the sample user jumping to the candidate work.

And 310, training the original model according to the target work and prediction result of the sample user jumping the slot to obtain a work prediction model.

According to the technical scheme of the embodiment of the application, the static attribute characteristic representation, the dynamic attribute characteristic representation and the time coding representation of the sample resume are determined according to the resume information of the sample user, the resume characteristic representation of the sample user is used as the resume characteristic representation, the dimension division of the resume characteristic of the sample user is finer, the sample user resume characteristic representation is used for representing a work prediction model obtained by training, the static attribute characteristic representation of the sample resume representing the user attribute and the dynamic attribute characteristic representation of the sample resume representing the work experience of the user are considered, and the accuracy of a work model prediction result obtained by training is greatly improved.

Fig. 4A is a flowchart of another method for establishing a work prediction model according to an embodiment of the present application. Fig. 4B is a schematic diagram illustrating an operation principle of a predictive neural network of an original model according to an embodiment of the present application. On the basis of the above embodiment, the present embodiment further optimizes the process of predicting the probability of the sample user jumping to the candidate job, and specifically includes the following steps:

s401, according to the sample user resume information, determining sample user resume feature representation and target work of sample user slot skipping.

S402, inputting the sample user resume feature representation into a time sequence neural network in the original model to obtain a user working experience representation.

And S403, inputting the user work experience representation into the collaborative neural network in the original model to obtain the user work characteristic representation.

S404, inputting the feature representation, the groove jumping time and the user work feature representation of the candidate work into a prediction neural network in the original model, and predicting the probability from groove jumping of the sample user to the candidate work.

Optionally, in the process of predicting the candidate working probability of the sample user for slot skipping, slot skipping time is introduced, and the slot skipping time can be predicted by the prediction neural network of the original model, or input by the user according to the requirement of the user.

Optionally, the predictive neural network in the embodiment of the present application includes a slot jumping company prediction sub-network and/or a slot jumping position prediction sub-network. The user work characteristic representation obtained by S403 may comprise a user company characteristic representation and/or a user post characteristic representation. When the user working characteristic representation comprises the user company characteristic representation, the prediction neural network comprises a groove jumping company prediction sub-network, and the prediction neural network can predict the probability of the user jumping to each candidate company; when the user working characteristic representation comprises a user position characteristic representation, the prediction neural network comprises a groove jumping position prediction sub-network, and the prediction neural network can predict the probability of the user jumping to each candidate position. Specifically, the step may include at least one of the following two cases:

and in case one, inputting the feature representation of the candidate company in the feature representation of the candidate work, the slot jumping time and the user company feature representation in the user work feature representation into a slot jumping company prediction sub-network, and predicting the probability of the sample user jumping to the candidate company.

Specifically, as shown in fig. 4B, in this case, the prediction neural network 12 of the original model 1 includes a groove jump company prediction subnetwork 122, and the groove jump time D may be first obtained_iT+1Coding into a time-hopping coded representation d_iT+1(the specific encoding process has been introduced in the above embodiment, and is not described herein), and then input into a layer of fully-connected network according to formula e_iT+1＝Wd_iT+1+ b obtaining a new vector e after mapping_iT+1Wherein W and b are weight coefficients and bias vectors of the fully connected network. Then the user company characteristics are expressed u_iTAnd e_iT+1Adding to obtain the final hidden variables of the dynamic user companyAnd finally, predicting the probability of the sample user jumping to the candidate company according to the following formula (3) by using a normalized index (softmax) function.

Wherein, C_ij,T+1For the predicted probability of a sample user i jumping to a candidate company j,as a transpose of the feature representation of candidate company j,company hidden variables for dynamic users.

And secondly, inputting the feature representation of the candidate position in the feature representation of the candidate work, the slot jumping time and the user position feature representation in the user work feature representation into a slot jumping position prediction sub-network, and predicting the probability of the sample user jumping to the candidate position.

Specifically, as shown in fig. 4B, in this case, the predictive neural network 12 of the original model 1 includes a slot jump position prediction subnetwork 123, and a slot jump time D may be first obtained_iT+1Coding into a time-hopping coded representation d_iT+1(the specific encoding process has already been introduced in the above embodiment, and is not described herein), and then is input to a layer of fully-connected network according to the formula e'_iT+1＝W′d_iT+1+ b ' to get the new vector e ' after mapping '_iT+1Wherein, W 'and b' are weight coefficients and bias vectors of the fully-connected network. Then representing the user position characteristics w_iTAnd e'_iT+1Adding to obtain the final dynamic user post hidden variableAnd finally, predicting the probability of the sample user jumping to the candidate position according to the following formula (4) by utilizing a normalized index (softmax) function.

Wherein p is_ik,T+1For the predicted probability of a sample user i jumping to the candidate position k,as a transpose of the feature representation of candidate position k,company hidden variables for dynamic users.

S405, training the original model according to the target work and prediction result of the sample user jumping the slot to obtain a work prediction model.

Optionally, in this embodiment of the present application, a method for determining the feature representation of the candidate company in the feature representation of the candidate job may be to encode attribute information of each candidate company, such as a company name, an address, a telephone profile, and the like, into a vector or matrix representation, and input the vector or matrix representation into a multi-layer perceptual network that extracts the feature representation of the company, so as to obtain the feature representation of each candidate company. Since the position information does not change with time, the method for determining the feature representation of the candidate positions in the embodiment of the present application may set a hidden variable feature for each candidate position in advance as the feature representation of each candidate position. The same or different characteristic expressions may be used for the same post without using a company, and this embodiment is not limited to this.

According to the technical scheme of the embodiment of the application, when the probability from the sample user to the candidate work is predicted by constructing the prediction neural network of the trained original model, a time factor of the groove jump is introduced, and the probability from the sample user to the candidate work is predicted by combining the feature representation of the candidate work, the groove jump time and the user work feature representation. The corresponding dynamic work recommendation result can be given according to the dynamically changed slot skipping time, so that the work recommended to the user is more accurate.

Optionally, if the two situations predict the slot jumping position and the slot jumping time required by the slot jumping company are obtained by predicting the neural network itself, the predicting neural network further includes: a slot hop time prediction subnetwork; s403, obtaining the user working characteristic representation and the user time characteristic representation which needs to be included. Before this step, the method also includes inputting the user time characteristic representation in the user work characteristic representation into the slot jump time prediction sub-network, and predicting the slot jump time. Specifically, as shown in fig. 4B, in this case, the predictive neural network 12 of the original model 1 includes a groove-skipping time predictive self-network 121, where x can be represented for the user time characteristics_iTPerforming final optimal time-to-slot prediction through a layer of fully-connected neural network, wherein the spliced layer of fully-connected neural network is equivalent to the pair x_iTAnd performing logistic regression. Compared with the prior art, the working prediction model trained in the embodiment of the application can accurately predict the time of the user for jumping the groove on the basis of predicting the company and the post of jumping the groove, the prediction result is richer, and the time of jumping the groove also provides guarantee for the follow-up accurate prediction of the company and the post of jumping the groove.

FIG. 5A is a flowchart of a work recommendation method provided in accordance with an embodiment of the present application; FIG. 5B is a schematic diagram of a structure of a working prediction model provided according to an embodiment of the present application; the embodiment is suitable for the case that the work prediction model established in any one of the above embodiments is deployed in a work recommendation system to realize the recommendation of work for a target user. The embodiment can be executed by an electronic device where a work recommendation system is located, and the electronic device is configured with a work prediction model and a work recommendation device, and the device can be implemented by software and/or hardware. As shown in fig. 5A-5B, the method includes:

s501, according to the resume information of the target user, determining the resume characteristic representation of the target user.

Optionally, the target user in the embodiment of the present application may be a job seeker user. The step can be that resume information of the target user is obtained, and then according to the resume information of the target user and according to the training work prediction model, a method similar to the sample user resume feature representation is determined according to the sample user resume information, so as to determine the target user resume feature representation. Optionally, the specific execution process may be to determine attribute information of the target user, work attribute information of at least one piece of work of the target user, and a work duration of the piece of work according to the resume information of the target user; determining static attribute feature representation of the target resume according to the attribute information of the target user; determining dynamic attribute feature representation of the target resume according to at least one piece of work attribute information of the target user; coding at least one part of working time of a target user, and determining target time coding expression; and taking the static attribute feature representation, the dynamic attribute feature representation and the target time coding representation of the target resume as the feature representation of the target user resume. It should be noted that the process is similar to the process of determining sample user resume feature representation and target work of sample user jumping from slots according to sample user resume information in the above embodiment, and only the sample user resume in the above embodiment needs to be replaced by a target user resume, and the specific determination process is the same and is not described herein again. In the embodiment of the application, in the process of pre-feature extraction of the resume information of the target user, the dimension division of the extracted resume features of the target user is more precise, the predicted slot jump work is represented by the resume features of the target user, not only the static attribute feature representation of the target resume representing the attributes of the user is considered, but also the dynamic attribute feature representation of the target resume representing the work experience of the user is considered, and the accuracy of the prediction result of the work prediction model is greatly improved.

For example, assuming that the work history record of the resume of the target user has two works, the resume feature representation of the target user determined in this step includes a static attribute feature representation F of the target resume_isFirst-copy work target resume dynamic attribute feature representation F_i1Duration of operation of the first copy D_i1Corresponding target time-coded representation d_i1Second job's target resume dynamic Attribute feature representation F_i2And a working duration D of the second job_i2Corresponding target time-coded representation d_i2。

And S502, inputting the resume feature representation of the target user into the work prediction model to obtain the probability of the target user jumping to the candidate work.

Optional target user resume features determined in S501 include: the method comprises the steps of target resume static attribute feature representation, target resume dynamic attribute feature representation of each job in the target user job experience and target time coding representation. In the step, a method similar to the method for inputting the resume feature representation of the target user into the original model to be trained when the work prediction model is trained can be adopted to input the resume feature of the target user into the trained work prediction model, and the work prediction model analyzes the input feature representation of the target user according to an algorithm during training to determine the probability of the target user jumping to each candidate work.

Specifically, the initial transfer parameter may be obtained according to the static attribute feature representation of the target resume; taking the initial transmission parameters as transmission input of a first sub time sequence network of a time sequence neural network in the work prediction model; and sequentially taking the dynamic attribute feature representation and the target time coding representation of the target resume of each job and the static attribute feature representation of the target resume as the parameter input of a sub-time sequence network in the time sequence neural network. The process is similar to the process of inputting the sample user resume feature representation into the time-series neural network in the original model introduced in the above embodiment, and is not repeated in this embodiment. Each sub-time sequence network analyzes the transmission input and the parameter input to obtain two outputs, wherein one output is a user work experience representation which needs to be transmitted to a corresponding sub-perception network of the sub-time sequence network in the cooperative neural network, and the other output is a transmission parameter which is used as the parameter input of the next sub-time sequence network. Each sub-perception network in the collaborative neural network analyzes the received user work experience representation to determine a user work characteristic representation, and optionally, the user work characteristic representation includes: at least one of a user time characterization, a user company characterization, and a user post characterization. When predicting the next time slot skipping work for the target user, the embodiment of the application may use the user work feature representation output by the sub-perception network corresponding to the last sub-timing network in the time sequence arrangement as the input of the prediction neural network, so that the prediction network predicts the probability of the target user from the current work slot skipping to each candidate work according to the input user work feature representation corresponding to the current work. Optionally, the slot jump time prediction sub-network in the neural network predicts the slot jump time according to the user time characteristic representation. The slot jumping company prediction sub-network predicts the probability of the target user jumping to each candidate company according to the feature representation of the candidate company in the feature representation of the candidate work, the slot jumping time and the user company feature representation in the user work feature representation.

For example, when the work experience record of the resume of the target user has two works, this step may represent the static attribute feature of the target resume in the resume feature representation of the target user as F_isMapped initial transfer parametersAs a pass-through input for LSTM-1 in FIG. 5B, the first active target resume dynamic attribute feature in the target user resume feature representation is represented F_i1Duration of operation of the first copy D_i1Corresponding target time-coded representation d_i1And target resume static attribute feature representation F_isAs the parameter input of LSTM-1, the transfer parameter of LSTM-1 output is outputAs the transfer input of LSTM-2, the dynamic attribute feature representation F of the target resume of the second work in the resume feature representation of the target user_i2The working time length D of the second work_i2Corresponding target time-coded representation d_i2And target resume static attribute feature representation F_isAs a parameter input for LSTM-2. At this time, LSTM-1 and LSTM-2 will be processed based on the input parameters according to the algorithm of the LSTM training work prediction model in the above embodiment, and respectively output the user work experience representation corresponding to the first workUser work experience representation corresponding to a second jobThe embodiment of the application can be used for representing the user work experience corresponding to the second workInputting into MLP-2 corresponding to LSTM-2, in order to reduce power consumption, LSTM-1 may not represent user work experience corresponding to the first workTransmitting to MLP-1, only the output transfer parametersInput deviceTo LSTM-2. MLP-2 will process according to the algorithm when training MLP in the prediction model, get user's time characteristic to express x_i2User company feature representation u_i2And user position feature representation w_i2Input to the predictive neural network 52, the time to slot jump prediction subnetwork 521 in the predictive neural network 52 will predict the slot time according to x_i2Next slot jump time Di3 is predicted. The jumping company prediction subnetwork 522 will predict the candidate company's feature representation v from the candidate job's feature representations_j2、D_i3And u_i2The probability that the target user i jumps from the second job (i.e., the current job) to each candidate company j is predicted. The jumping slot position prediction subnetwork 523 will predict the characteristic representation q of the candidate position from the characteristic representations q of the candidate positions in the characteristic representations of the candidate operation_k、D_i3And w_i2And predicting the probability of the target user i jumping from the second work (namely the current work) to each candidate position k.

In the embodiment of the present invention, when predicting the probability of the user jumping to a candidate job, the time of the jumping may be predicted by the time prediction subnetwork 521 in the prediction neural network 52, or may be input according to the actual demand of the target user, which is not limited in this embodiment.

And S503, recommending work for the target user according to the probability of the candidate work of the user jumping the slot.

Optionally, in this step, when recommending work for the target user according to the probability of the user jumping slot candidate work, one or more pieces with higher probability may be selected from the candidate works as the work eventually recommended for the target user.

According to the technical scheme of the embodiment of the application, the work prediction model is constructed and trained by adopting any one of the embodiments. And determining the resume characteristic representation of the target user according to the resume information of the target user, inputting the resume characteristic representation of the target user into the work prediction model, obtaining the probability of the target user jumping from the slot to each candidate work, and determining the work recommended for the target user according to the probability of the candidate work after the target user jumps from the slot to the next candidate work. The working prediction model is trained on resume information of a large number of different sample users, so that when the working prediction model is used for predicting a certain user on a groove jumping working line in the subsequent process, the probability from the groove jumping of the target user to each candidate work can be determined by cooperating with the selection of other users similar to the working characteristic representation of the target user in the groove jumping process. The accuracy of the prediction result is greatly improved while work recommendation based on the personalized requirements of the user can be realized.

Fig. 6 is a schematic structural diagram of an apparatus for building a work prediction model according to an embodiment of the present application. The method is suitable for the condition of constructing and training the neural network model capable of executing the work prediction task, and the device can realize the method for establishing the work prediction model in any embodiment of the application.

The apparatus 600 specifically includes the following:

the sample preprocessing module 601 is used for determining sample user resume feature representation and target work of sample user slot skipping according to the sample user resume information;

the model training module 602 is used for inputting the sample user resume feature representation into the time sequence neural network in the original model to obtain a user work experience representation, inputting the user work experience representation into the collaborative neural network in the original model to obtain a user work feature representation, inputting the feature representation of candidate work and the user work feature representation into the prediction neural network in the original model, and predicting the probability of the sample user jumping to the candidate work; and training the original model according to the target work and prediction results of the sample user slot jumping to obtain a work prediction model.

According to the technical scheme, an original model comprising a time sequence neural network, a cooperative perception network and a prediction neural network is constructed, user resume characteristic representation determined based on sample user resume information is input to the time sequence neural network of the original model, user working experience representation is input to the cooperative neural network, user working characteristic representation is input to the prediction neural network, the probability that a sample user jumps to candidate work is predicted, and the original model is trained by combining target work of actual groove jumping of the sample user to obtain a working prediction model. The working prediction model is trained on resume information of a large number of different sample users, so that when the working prediction model is used for predicting a certain user on a groove jumping working line in the subsequent process, the probability from the groove jumping of the target user to each candidate work can be determined by cooperating with the selection of other users similar to the working characteristic representation of the target user in the groove jumping process. The accuracy of the prediction result is greatly improved while work recommendation based on the personalized requirements of the user can be realized.

Furthermore, sub time sequence networks in the time sequence neural network correspond to sub perception networks in the cooperative neural network one to one;

the first output end of the sub-time sequence network is connected with the next sub-time sequence network;

and the second output end of the sub time sequence network is connected with the corresponding sub perception network.

Further, the sub-sensing network comprises at least one of a first sub-sensing unit, a second sub-sensing unit and a third sub-sensing unit; the input ends of the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are all connected with the second output end of the corresponding sub-timing network, and the output ends of the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are all connected with the input end of the prediction neural network;

the first sub-sensing unit, the second sub-sensing unit and the third sub-sensing unit are respectively used for outputting user time characteristic representation, user company characteristic representation and user post characteristic representation in the user work characteristic representation.

Further, different sub-aware networks in the collaborative neural network share model parameters.

Further, the sub-timing network is a long-short term memory network LSTM; the sub-perception network is a multi-layer perception machine MLP.

Further, the sample preprocessing module 601 includes:

the information determining unit is used for determining the attribute information of the sample user, the working attribute information of at least one part of work of the sample user and the working duration of the part of work according to the resume information of the sample user;

the static characteristic determining unit is used for determining the static attribute characteristic representation of the sample resume according to the sample user attribute information;

the dynamic characteristic determining unit is used for determining dynamic attribute characteristic representation of the sample resume according to the working attribute information of at least one job of the sample user;

the time coding unit is used for coding the working time of at least one part of work of a sample user and determining the sample time coding expression;

and the characteristic integration unit is used for taking the sample resume static attribute characteristic representation, the sample resume dynamic attribute characteristic representation and the sample time coding representation as the sample user resume characteristic representation.

Further, the model training module 602 includes a first data input unit for inputting the sample user resume feature representation into a time-series neural network in the original model; the first data input unit is specifically configured to:

obtaining an initial transmission parameter according to the static attribute feature representation of the sample resume;

taking the initial transmission parameters as transmission input of a first sub time sequence network of the time sequence neural network in the original model;

and sequentially taking the dynamic attribute feature representation and the time code representation of the sample resume of each job and the static attribute feature representation of the sample resume as the parameter input of a sub-time sequence network in the time sequence neural network.

Further, the model training module 602 includes a second data input unit for inputting the feature representation of the candidate job and the user job feature representation into a prediction neural network in the original model, predicting a probability of the sample user jumping to the candidate job; the second data input unit is specifically configured to: and inputting the feature representation of the candidate work, the slot jumping time and the user work feature representation into a prediction neural network in the original model, and predicting the probability of the sample user from slot jumping to the candidate work.

Further, the prediction neural network comprises a slot jumping company prediction sub-network and/or a slot jumping position prediction sub-network;

correspondingly, the second data input unit is specifically configured to:

inputting the feature representation of the candidate company in the feature representation of the candidate work, the slot jumping time and the user company feature representation in the user work feature representation into a slot jumping company prediction sub-network, and predicting the probability of the sample user from jumping to the candidate company; and/or the presence of a gas in the gas,

and inputting the feature representation of the candidate position in the feature representation of the candidate work, the slot jumping time and the user position feature representation in the user work feature representation into a slot jumping position prediction sub-network, and predicting the probability of the sample user jumping to the candidate position.

Further, the predicting neural network further includes: a slot hop time prediction subnetwork; accordingly, the model training module 602 further includes:

and the third data input unit is used for inputting the user time characteristic representation in the user work characteristic representation into the groove jumping time prediction sub-network to predict the groove jumping time.

Fig. 7 is a schematic structural diagram of a work recommendation device according to an embodiment of the present application. The embodiment is suitable for the case that the work prediction model established in any one of the above embodiments is deployed in a work recommendation system to realize the recommendation of work for a target user. The device can realize the work recommendation method of any embodiment of the application. The apparatus 700 specifically includes the following:

the user resume processing module 701 is used for determining the resume characteristic representation of the target user according to the resume information of the target user;

the work prediction module 702 is configured to input the target user resume feature representation into a work prediction model to obtain a probability that the target user jumps to the candidate work;

and the work recommending module 703 is configured to recommend work to the target user according to the probability of the candidate work for the user to jump the slot.

According to the technical scheme of the embodiment of the application, the work prediction model is constructed and trained by adopting any one of the embodiments. And determining the resume characteristic representation of the target user according to the resume information of the target user, inputting the resume characteristic representation of the target user into the work prediction model, obtaining the probability of the target user jumping from the slot to each candidate work, and determining the work recommended for the target user according to the probability of the candidate work after the target user jumps from the slot to the next candidate work. The working prediction model is trained on resume information of a large number of different sample users, so that when the working prediction model is used for predicting a certain user on a groove jumping working line in the subsequent process, the probability from the groove jumping of the target user to each candidate work can be determined by cooperating with the selection of other users similar to the working characteristic representation of the target user in the groove jumping process. The accuracy of the prediction result is greatly improved while work recommendation based on the personalized requirements of the user can be realized.

Further, the user resume processing module 701 is specifically configured to:

according to the resume information of the target user, determining attribute information of the target user, working attribute information of at least one part of work of the target user and working duration of the part of work;

determining static attribute feature representation of the target resume according to the attribute information of the target user;

determining dynamic attribute feature representation of the target resume according to at least one piece of work attribute information of the target user;

coding at least one part of working time of a target user, and determining target time coding expression;

and taking the static attribute feature representation, the dynamic attribute feature representation and the target time coding representation of the target resume as the feature representation of the target user resume.

Further, the work prediction module 702 is specifically configured to:

obtaining an initial transfer parameter according to the static attribute feature representation of the target resume;

taking the initial transmission parameters as transmission input of a first sub time sequence network of a time sequence neural network in the work prediction model;

and sequentially taking the dynamic attribute feature representation and the target time coding representation of the target resume of each job and the static attribute feature representation of the target resume as the parameter input of a sub-time sequence network in the time sequence neural network.

According to an embodiment of the present application, an electronic device and a readable storage medium are also provided.

Fig. 8 is a block diagram of an electronic device according to an embodiment of the present application, illustrating a method for building a work prediction model or a method for recommending a work. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the present application that are described and/or claimed herein.

As shown in fig. 8, the electronic apparatus includes: one or more processors 801, memory 802, and interfaces for connecting the various components, including a high speed interface and a low speed interface. The various components are interconnected using different buses and may be mounted on a common motherboard or in other manners as desired. The processor may process instructions for execution within the electronic device, including instructions stored in or on the memory to display graphical information of a GUI on an external input/output apparatus (such as a display device coupled to the interface). In other embodiments, multiple processors and/or multiple buses may be used, along with multiple memories and multiple memories, as desired. Also, multiple electronic devices may be connected, with each device providing portions of the necessary operations (e.g., as a server array, a group of blade servers, or a multi-processor system). Fig. 8 illustrates an example of a processor 801.

The memory 802 is a non-transitory computer readable storage medium as provided herein. The memory stores instructions executable by the at least one processor to cause the at least one processor to perform a method for building a work prediction model or a method for recommending work provided herein. The non-transitory computer-readable storage medium of the present application stores computer instructions for causing a computer to execute the work prediction model building method or the work recommendation method provided by the present application.

The memory 802 is a non-transitory computer readable storage medium, and can be used to store non-transitory software programs, non-transitory computer executable programs, and modules, such as program instructions/modules corresponding to the work prediction model building method or the work recommendation method in the embodiments of the present application (for example, the sample preprocessing module 601 and the model training module 602 shown in fig. 6, or the user resume processing module 701, the work prediction module 702, and the work recommendation module 703 shown in fig. 7). The processor 801 executes various functional applications of the server and data processing, i.e., a method of establishing a work prediction model or a work recommendation method in the above-described method embodiments, by executing non-transitory software programs, instructions, and modules stored in the memory 802.

The memory 802 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the electronic device of the establishment method of the work prediction model or the work recommendation method, or the like. Further, the memory 802 may include high speed random access memory and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory 802 may optionally include a memory remotely disposed from the processor 801, and these remote memories may be connected to an electronic device of a work prediction model creation method or a work recommendation method through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The electronic device of the work prediction model establishing method or the work recommendation method may further include: an input device 803 and an output device 804. The processor 801, the memory 802, the input device 803, and the output device 804 may be connected by a bus or other means, and are exemplified by a bus in fig. 8.

The input device 803 may receive input numeric or character information and generate key signal inputs related to user settings and function control of the electronic device of the work prediction model establishment method or the work recommendation method, such as an input device of a touch screen, a keypad, a mouse, a track pad, a touch pad, a pointing stick, one or more mouse buttons, a track ball, a joystick, or the like. The output devices 804 may include a display device, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibrating motors), among others. The display device may include, but is not limited to, a Liquid Crystal Display (LCD), a Light Emitting Diode (LED) display, and a plasma display. In some implementations, the display device can be a touch screen.

Various implementations of the systems and techniques described here can be realized in digital electronic circuitry, integrated circuitry, application specific ASICs (application specific integrated circuits), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

These computer programs (also known as programs, software applications, or code) include machine instructions for a programmable processor, and may be implemented using high-level procedural and/or object-oriented programming languages, and/or assembly/machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and/or device (e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs)) used to provide machine instructions and/or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal used to provide machine instructions and/or data to a programmable processor.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), and the Internet.

The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

According to the technical scheme of the embodiment of the application, an original model comprising a time sequence neural network, a cooperative perception network and a prediction neural network is constructed, user resume characteristic representation determined based on sample user resume information is input to the time sequence neural network of the original model, user working experience representation is obtained and input to the cooperative neural network, user working characteristic representation is obtained and input to the prediction neural network, the probability that a sample user jumps to candidate work is predicted, and then the original model is trained by combining with target work of actual groove jumping of the sample user, so that a working prediction model is obtained. The working prediction model is trained on resume information of a large number of different sample users, so that when the working prediction model is used for predicting a certain user on a groove jumping working line in the subsequent process, the probability from the groove jumping of the target user to each candidate work can be determined by cooperating with the selection of other users similar to the working characteristic representation of the target user in the groove jumping process. The accuracy of the prediction result is greatly improved while work recommendation based on the personalized requirements of the user can be realized.

It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present application may be executed in parallel, sequentially, or in different orders, and the present invention is not limited thereto as long as the desired results of the technical solutions disclosed in the present application can be achieved.

The above-described embodiments should not be construed as limiting the scope of the present application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.