阅读说明：本技术 基于机器学习的医疗风险信息推送方法及装置 (Medical risk information pushing method and device based on machine learning ) 是由 李映雪 熊昊 李响 于 2021-08-31 设计创作，主要内容包括：本申请提供了一种基于机器学习的医疗风险信息推送方法及装置,涉及人工智能及数字医疗领域,主要目的在于改善现有患病风险发生率升高,以及就医数据处理准确性降低的问题。包括：获取预设采集设备的全部位点信息；基于筛选处理模型从全部位点信息中筛选目标测试位点,提取与目标测试位点匹配的医疗检测数据；基于已完成模型训练的风险预测模型对医疗检测数据进行风险预测处理,并解析风险预测处理后得到的医疗预测结果中与目标病症信息匹配的第一过程信息；若已建立的目标病症用户画像数据库中各用户画像数据所对应的第二过程信息与第一过程信息的相似度超过预设相似度阈值,则获取用户画像数据中与医疗预测结果匹配的医疗风险信息,并进行推送。(The application provides a medical risk information pushing method and device based on machine learning, relates to the field of artificial intelligence and digital medical treatment, and mainly aims to solve the problems that the existing risk incidence rate of illness is increased, and the accuracy of medical data processing is reduced. The method comprises the following steps: acquiring all site information of preset acquisition equipment; screening target test sites from all site information based on a screening processing model, and extracting medical detection data matched with the target test sites; performing risk prediction processing on the medical detection data based on a risk prediction model which is trained by the model, and analyzing first process information matched with target disease information in a medical prediction result obtained after the risk prediction processing; and if the similarity between the first process information and the second process information corresponding to each user portrait data in the established target disease user portrait database exceeds a preset similarity threshold, acquiring medical risk information matched with the medical prediction result in the user portrait data, and pushing the medical risk information.)

1. A medical risk information pushing method based on machine learning is characterized by comprising the following steps:

acquiring all site information of preset acquisition equipment;

screening target test sites from all the site information based on a screening processing model, and extracting medical detection data matched with the target test sites, wherein the screening processing model is obtained by adjusting model hyper-parameters based on model evaluation indexes of a test validation set and completing training;

performing risk prediction processing on the medical detection data based on a risk prediction model which is trained by the model, and analyzing first process information matched with target disease information in a medical prediction result obtained after the risk prediction processing, wherein the first process information is used for representing disease time, treatment information and a medical stage matched with the target disease process information;

and if the similarity between the second process information corresponding to each user portrait data in the established target disease user portrait database and the first process information exceeds a preset similarity threshold, acquiring medical risk information matched with the medical prediction result in the user portrait data, and pushing the medical risk information.

2. The method of claim 1, wherein before obtaining all location information of the predetermined collection device, the method further comprises:

acquiring a site sample test feature set, and determining a first preset number of principal component features;

sorting the magnitude sequence through the coefficient absolute values of the main component characteristics, screening the site information of a second preset number, and calculating a model evaluation index based on the site sample test feature set;

and configuring and adjusting model hyperparameters through the model evaluation indexes, and training a screening processing model by combining the screened site information and the model hyperparameters.

3. The method of claim 2, wherein said obtaining a site sample test feature set, and wherein determining a first preset number of principal component features comprises:

extracting at least one site position information in the site sample test set and a site mark for marking the integrity of the site information;

selecting the position point information corresponding to at least two position point identifications in the position point sample test set according to row and column units in sequence, and counting the number of the selected position point information;

and determining the position information corresponding to the first preset number as the principal component characteristics.

4. The method of claim 1, wherein prior to performing risk prediction processing on the medical examination data based on the model-trained risk prediction model, the method further comprises:

and constructing a convolutional neural network model, and performing model training on the convolutional neural network model based on a medical detection data sample set to obtain the risk prediction model, wherein the risk pre-model is used for configuring the hierarchical weight of the convolutional neural network model based on the number of test site samples in the medical detection data sample set to complete iterative training.

5. The method of claim 1, further comprising:

acquiring user basic data matched with a target disease condition in a user medical database, and disease condition time, treatment time and a medical stage associated with the target disease condition;

and establishing a target disease user image database based on the user basic data, the disease time, the treatment time and the medical stage.

6. The method of claim 5, wherein after establishing a target medical condition user profile database based on the user base data, the medical condition time, the visit time, and the medical treatment session, the method further comprises:

acquiring diagnosis information generated in the visiting process of each user in the target disease user picture database, current symptom information and expected symptom information matched with the diagnosis information;

and combining the diagnosis information, the current symptom information and the expected symptom information to generate medical risk information matched with different medical prediction results.

7. The method according to any one of claims 1 to 6, wherein after acquiring and pushing the medical risk information matching the medical prediction result in the user profile data, the method further comprises:

recording the receiving times of the medical risk information, and collecting medical operations generated by each user based on the medical risk information;

and updating the preset similarity threshold according to the similarity of the medical operation and the visit information of the first process information.

8. A medical risk information pushing device based on machine learning is characterized by comprising:

the acquisition module is used for acquiring all the site information of the preset acquisition equipment;

the screening module is used for screening target test sites from all the site information based on a screening processing model and extracting medical detection data matched with the target test sites, and the screening processing model is obtained by adjusting model hyper-parameters based on model evaluation indexes of a test verification set and completing training;

the analysis module is used for carrying out risk prediction processing on the medical detection data based on a risk prediction model which is trained by the model, and analyzing first process information matched with target disease information in a medical prediction result obtained after the risk prediction processing, wherein the first process information is used for representing disease time, treatment information and a medical stage matched with the target disease information;

and the pushing module is used for acquiring medical risk information matched with the medical prediction result in the user portrait data and pushing the medical risk information if the similarity between the first process information and the second process information corresponding to each user portrait data in the established target disease user portrait database exceeds a preset similarity threshold.

9. A storage medium having at least one executable instruction stored therein, wherein the executable instruction causes a processor to execute operations corresponding to the machine learning-based medical risk information pushing method according to any one of claims 1-7.

10. A computer device, comprising: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus;

the memory is used for storing at least one executable instruction, and the executable instruction causes the processor to execute the operation corresponding to the medical risk information pushing method based on machine learning in any one of claims 1-7.

Technical Field

The application relates to the technical field of artificial intelligence and digital medical treatment, in particular to a medical risk information pushing method and device based on machine learning.

Background

With the rapid development of intelligent medical treatment, more and more patients choose to seek medical treatment or medical consultation in a digital medical treatment mode. Wherein, in order to realize the timely doctor seeing of patient, can carry out the propelling movement with the relevant medical risk that the patient got a doctor in-process and produced through the mode of medical risk information propelling movement, the serious production risk of the patient's state of an illness that significantly reduces.

Currently, the existing medical risk information is only pushed according to the characteristics of time, age, disease and the like, for example, the diabetic kidney risk information is pushed to the diabetic patient every month, so that the diabetic patient can timely pay attention to the self condition for medical treatment. However, only by using a specific characteristic as a pushing basis, information cannot be pushed timely and flexibly according to the self condition of a patient, so that the incidence rate of the disease risk is greatly increased, and the effectiveness of information pushing is reduced, thereby affecting the accuracy of processing medical data in an intelligent medical system.

Disclosure of Invention

In view of this, the present application provides a medical risk information pushing method and device based on machine learning, and mainly aims to solve the technical problems that the existing method cannot timely and flexibly push information according to the self condition of a patient, the effectiveness of information pushing is reduced, the incidence rate of illness risk is increased, and the accuracy of medical data processing in an intelligent medical system is reduced.

According to one aspect of the application, a medical risk information pushing method based on machine learning is provided, and comprises the following steps:

acquiring all site information of preset acquisition equipment;

screening target test sites from all the site information based on a screening processing model, and extracting medical detection data matched with the target test sites, wherein the screening processing model is obtained by adjusting model hyper-parameters based on model evaluation indexes of a test validation set and completing training;

performing risk prediction processing on the medical detection data based on a risk prediction model which is trained by the model, and analyzing first process information matched with target disease information in a medical prediction result obtained after the risk prediction processing, wherein the first process information is used for representing disease time, treatment information and a medical stage matched with the target disease process information;

and if the similarity between the second process information corresponding to each user portrait data in the established target disease user portrait database and the first process information exceeds a preset similarity threshold, acquiring medical risk information matched with the medical prediction result in the user portrait data, and pushing the medical risk information.

Preferably, before the acquiring of all the location information of the preset acquisition device, the method further includes:

acquiring a site sample test feature set, and determining a first preset number of principal component features;

sorting the magnitude sequence through the coefficient absolute values of the main component characteristics, screening the site information of a second preset number, and calculating a model evaluation index based on the site sample test feature set;

and configuring and adjusting model hyperparameters through the model evaluation indexes, and training a screening processing model by combining the screened site information and the model hyperparameters.

Preferably, the obtaining a site sample test feature set, and determining a first preset number of principal component features includes:

extracting at least one site position information in the site sample test set and a site mark for marking the integrity of the site information;

selecting the position point information corresponding to at least two position point identifications in the position point sample test set according to row and column units in sequence, and counting the number of the selected position point information;

and determining the position information corresponding to the first preset number as the principal component characteristics.

Preferably, before the risk prediction processing is performed on the medical examination data based on the risk prediction model trained by the completed model, the method further includes:

and constructing a convolutional neural network model, and performing model training on the convolutional neural network model based on a medical detection data sample set to obtain the risk prediction model, wherein the risk pre-model is used for configuring the hierarchical weight of the convolutional neural network model based on the number of test site samples in the medical detection data sample set to complete iterative training.

Preferably, the method further comprises:

acquiring user basic data matched with a target disease condition in a user medical database, and disease condition time, treatment time and a medical stage associated with the target disease condition;

and establishing a target disease user image database based on the user basic data, the disease time, the treatment time and the medical stage.

Preferably, after the target disease user image database is established based on the user basic data, the disease time, the treatment time and the medical treatment stage, the method further comprises:

acquiring diagnosis information generated in the visiting process of each user in the target disease user picture database, current symptom information and expected symptom information matched with the diagnosis information;

and combining the diagnosis information, the current symptom information and the expected symptom information to generate medical risk information matched with different medical prediction results.

Preferably, after acquiring and pushing medical risk information matching the medical prediction result in the user profile data, the method further includes:

recording the receiving times of the medical risk information, and collecting medical operations generated by each user based on the medical risk information;

and updating the preset similarity threshold according to the similarity of the medical operation and the visit information of the first process information.

According to another aspect of the present application, there is provided a medical risk information pushing apparatus based on machine learning, including:

the first acquisition module is used for acquiring all the site information of the preset acquisition equipment;

the screening module is used for screening target test sites from all the site information based on a screening processing model and extracting medical detection data matched with the target test sites, and the screening processing model is obtained by adjusting model hyper-parameters based on model evaluation indexes of a test verification set and completing training;

the analysis module is used for carrying out risk prediction processing on the medical detection data based on a risk prediction model which is trained by the model, and analyzing first process information matched with target disease information in a medical prediction result obtained after the risk prediction processing, wherein the first process information is used for representing disease time, treatment information and a medical stage matched with the target disease information;

and the pushing module is used for acquiring medical risk information matched with the medical prediction result in the user portrait data and pushing the medical risk information if the similarity between the first process information and the second process information corresponding to each user portrait data in the established target disease user portrait database exceeds a preset similarity threshold.

Preferably, before the first obtaining module, the apparatus further includes:

the determining module is used for acquiring a site sample test feature set and determining a first preset number of principal component features;

the calculation module is used for carrying out magnitude sequence sequencing through the coefficient absolute values of all the principal component characteristics, screening the site information of a second preset number, and calculating a model evaluation index based on the site sample test feature set;

and the training module is used for configuring and adjusting model hyperparameters through the model evaluation indexes and training the screening processing model by combining the screened site information and the model hyperparameters.

Preferably, the determining module comprises:

the extraction unit is used for extracting at least one site position information in the site sample test set and a site mark for marking the integrity of the site information;

the statistical unit is used for selecting the site position information corresponding to at least two site identifications in the site sample test set according to row and column units in sequence and counting the number of the selected site position information;

and the determining unit is used for determining the position information corresponding to the first preset number as the principal component characteristic based on the ratio of the number to the number of the collection length units of the preset collection equipment.

Preferably, before the parsing module, the apparatus further includes:

the risk prediction model comprises a construction module and a risk prediction module, wherein the construction module is used for constructing a convolutional neural network model and carrying out model training on the convolutional neural network model based on a medical detection data sample set to obtain the risk prediction model, and the risk prediction model is used for completing iterative training by configuring the hierarchical weight of the convolutional neural network model based on the number of test site samples in the medical detection data sample set.

Preferably, the apparatus further comprises:

the second acquisition module is used for acquiring user basic data matched with a target disease state in a user medical database, and disease state time, treatment time and a medical stage which are associated with the target disease state;

and the establishing module is used for establishing a target disease user image database based on the user basic data, the disease time, the treatment time and the medical stage.

Preferably, after the establishing module, the apparatus further includes:

the third acquisition module is used for acquiring diagnosis information generated in the visiting process of each user in the target disease user image database, and current symptom information and expected symptom information matched with the diagnosis information;

and the generation module is used for combining the diagnosis information, the current symptom information and the expected symptom information to generate medical risk information matched with different medical prediction results.

Preferably, after the pushing module, the apparatus further includes:

the recording module is used for recording the receiving times of the medical risk information and acquiring medical operations generated by each user based on the medical risk information;

and the updating module is used for updating the preset similarity threshold according to the similarity between the medical operation and the visit information of the first process information.

According to still another aspect of the present application, a storage medium is provided, where at least one executable instruction is stored, and the executable instruction causes a processor to perform operations corresponding to the machine learning-based medical risk information pushing method as described above.

According to yet another aspect of the present application, there is provided a computer device including: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus;

the memory is used for storing at least one executable instruction, and the executable instruction enables the processor to execute the operation corresponding to the medical risk information pushing method based on machine learning.

By means of the technical scheme, the technical scheme provided by the embodiment of the application at least has the following advantages:

the application provides a medical risk information pushing method and device based on machine learning. Compared with the prior art, the method and the device have the advantages that all the site information of the preset acquisition equipment is obtained; screening target test sites from all the site information based on a screening processing model, and extracting medical detection data matched with the target test sites, wherein the screening processing model is obtained by adjusting model hyper-parameters based on model evaluation indexes of a test validation set and completing training; performing risk prediction processing on the medical detection data based on a risk prediction model which is trained by the model, and analyzing first process information matched with target disease information in a medical prediction result obtained after the risk prediction processing, wherein the first process information is used for representing disease time, treatment information and a medical stage matched with the target disease process information; if the similarity between second process information corresponding to each user portrait data in the established target disease user portrait database and the first process information exceeds a preset similarity threshold, medical risk information matched with the medical prediction result in the user portrait data is obtained and pushed, the information can be pushed timely and flexibly according to the self condition of a patient, the effectiveness of information pushing is improved, the incidence rate of the illness risk is reduced, and meanwhile, the accuracy of processing medical data in the intelligent medical system is improved.

The foregoing description is only an overview of the technical solutions of the present application, and the present application can be implemented according to the content of the description in order to make the technical means of the present application more clearly understood, and the following detailed description of the present application is given in order to make the above and other objects, features, and advantages of the present application more clearly understandable.

Drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the application. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:

fig. 1 shows a flowchart of a medical risk information pushing method based on machine learning according to an embodiment of the present application;

fig. 2 shows a block diagram of a medical risk information pushing device based on machine learning according to an embodiment of the present application;

fig. 3 shows a schematic structural diagram of a computer device provided in an embodiment of the present application.

Detailed Description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

The embodiment of the application can acquire and process related data based on an artificial intelligence technology. Among them, Artificial Intelligence (AI) is a theory, method, technique and application system that simulates, extends and expands human Intelligence using a digital computer or a machine controlled by a digital computer, senses the environment, acquires knowledge and uses the knowledge to obtain the best result.

The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.

Based on this, in an embodiment, as shown in fig. 1, a medical risk information pushing method based on machine learning is provided, which is described by taking the method applied to computer devices such as a server as an example, where the server may be an independent server, or may be a cloud server that provides basic cloud computing services such as cloud service, cloud database, cloud computing, cloud function, cloud storage, Network service, cloud communication, middleware service, domain name service, security service, Content Delivery Network (CDN), and big data and artificial intelligence platform, such as an intelligent medical system, a digital medical platform, and the like. The method comprises the following steps:

101. and acquiring all the site information of the preset acquisition equipment.

Wherein, the collecting device is a medical detection device containing a chip, such as a blood collecting device, a urine detection device and the like. In the embodiment of the present application, a preset sugar kidney genetic information SNP detection device is taken as an example, and is a blood sampling device including a SNP chip, and the SNP chip can decompose genetic information in blood, that is, perform further data processing as detection data of SNP sites. In addition, since the SNP genetic information is recorded in each locus, all loci of the SNP detecting device preset with the sugarkidney genetic information are acquired for convenience of data processing.

It should be noted that, in the embodiment of the present application, acquiring all the sites is to perform data acquisition on blood sampling information of a current user, so as to push risk information to the user, for example, an offline user performs code scanning and blood sampling through a fixed detection device, and then pushes risk information obtained after analysis and matching to the user according to identity information recorded by the code scanning.

102. And screening target test sites from all site information based on a screening processing model, and extracting medical detection data matched with the target test sites.

The screening processing model is obtained by adjusting model hyper-parameters based on model evaluation indexes of the test validation set and completing training.

In the embodiment of the application, the target test site is obtained by performing prediction screening based on a screening processing model. And the screening processing model is obtained by adjusting the hyper-parameters in the model by using the model evaluation indexes of the test validation set to complete model training. The extraction method of the medical detection data comprises the steps of selecting a preset number of main components, sorting the main components according to the magnitude of coefficient absolute values of the main components, screening preset number of site data, combining the site data screened by the main components to obtain medical detection numerical values corresponding to the site data, and further utilizing the detection data to conduct prediction processing.

It should be noted that, because the space of the panel is limited, in general, the number of the test sites that can be accommodated in one panel is about 100, so that the test sites that are most favorable for the prediction effect and the number of the test sites is within 100 need to be selected from millions of test sites, so as to achieve the efficient performance of accurately predicting the medical test result. And (4) screening target test sites from all the sites, and performing prediction processing on medical detection data corresponding to the target test site data by combining the trained prediction model to obtain a medical prediction result. Further, the prediction model is used to perform prediction processing on target disease risk on the medical detection data corresponding to each target test site, so that the corresponding medical prediction result includes target disease risk distributions or risk states of different degrees, such as low-level risk, medium-level risk, and high-level risk, which may also be divided more finely, and the embodiment of the present application is not particularly limited. The prediction model may be a machine learning model such as a neural network model and a support vector machine model, and then prediction processing is performed to complete model training, which is not specifically limited in the embodiments of the present application.

103. And performing risk prediction processing on the medical detection data based on the risk prediction model which is trained by the model, and analyzing first process information matched with the target disease information in the medical prediction result obtained after the risk prediction processing.

Wherein the first process information is used for representing the disease time, the clinic information and the medical stage which are matched with the target disease process information.

In the embodiment of the present application, the target disease is targeted physical index monitoring performed by a user sensing physical abnormality, and thus obtained disease information, such as a diabetic nephropathy disease. The first process information is process information of a current user for risk prediction, such as a length of illness of the current user, treatment items that have been received, and medical stages divided for symptoms. Since the medical prediction result indicates that the user is at risk for the target disease, the course information matching the target disease information in the medical prediction result is further analyzed to recommend information to the user. The disease course information is used for representing time, treatment scheme and treatment stage of different target diseases, and can be matched with the target disease course timeline based on different risk states to obtain matched disease course information.

It should be noted that the target disease course timeline is generated based on a picture database of a large number of target disease patients, that is, all times of getting target diseases from a user picture database and corresponding treatment stages are obtained, and the time period of the target disease is divided according to at least 5 time stages to obtain at least 5 disease course stages, so as to match disease course information, including early stage, middle and later stages, late stage of severe illness, each disease course corresponds to different treatment stages, and also can be divided more finely in order to make the pushed information more accurate, and the embodiment of the present application is not limited specifically.

104. And if the similarity between the second process information corresponding to each user portrait data in the established target disease user portrait database and the first process information exceeds a preset similarity threshold, acquiring medical risk information matched with the medical prediction result in the user portrait data, and pushing the medical risk information.

In the embodiment of the application, the second process information is process information of a user in the target disease user image database, and the specific content may be similar to the first process information. In the embodiment of the application, the user images of all target disease state users are pre-established and stored in the target disease state user image database, so that when information is pushed, in order to improve the reliability and enable the users to timely and accurately seek medical advice, the user images with the disease course information similarity larger than the preset threshold value are searched in the established target disease state user image database. The method comprises the steps of firstly calculating the similarity between the course information of a current user and all the course information in a picture database, and if the similarity is larger than a preset threshold value, determining the corresponding user picture. Generally, the diagnosis and treatment information, the confirmed diagnosis information, the disease information and the like collected by the matched user are recorded in the determined user portrait, so that risk information expected to be generated at a certain stage (such as the late stage of the diabetes and nephropathy) relative to a target disease is extracted from the matched user portrait and pushed, and a user for risk prediction can know the body state of the user and effectively select a medical scheme.

For further explanation and limitation, in the embodiment of the present application, before acquiring all location information of the preset acquisition device, the method of the embodiment further includes: acquiring a site sample test feature set, and determining a first preset number of principal component features; sorting the magnitude sequence through the coefficient absolute values of the main component characteristics, screening the site information of a second preset number, and calculating a model evaluation index based on a site sample test feature set; and (4) configuring and adjusting model hyper-parameters through the model evaluation indexes, and training the screening processing model by combining the screened site information and the model hyper-parameters.

Specifically, the screening of the target test site can be divided into primary screening of characteristics and fine screening of characteristics. Wherein the preliminary screening of characteristics comprises: performing principal component analysis on the training set, and selecting the first K principal component characteristics (namely the principal component characteristics of the first preset number); sorting the coefficients of each principal component according to absolute values, selecting the first N sites, combining and removing the N sites to be combined together to be used as a candidate site set, setting the set to contain M sites (namely the site information of the second preset number), and then performing fine feature screening. Wherein, the fine feature screening specifically comprises the following steps: training on a training set by using a screening processing model based on M candidate sites, and then selecting Q sites with the highest feature importance ranking in the screening processing model, wherein Q is the maximum number of target detection sites which can be contained in one panel; the screening process model is trained on the training set using Q sites. The super-parameter adjustment is to adjust the super-parameters through model evaluation indexes of the models on the verification set, verify the model effect on the test set based on the finally selected site set and the trained screening processing model, and accordingly complete the training of the screening processing model.

In this embodiment of the present application, it is further preferable that the obtaining a site sample test feature set, and determining the first preset number of principal component features includes: extracting at least one site position information in the site sample test set and a site mark for marking the integrity of the site information; selecting the site position information corresponding to at least two site identifications in the site sample test set according to row and column units in sequence, and counting the number of the selected site position information; and determining the position information corresponding to the first preset number as the principal component characteristics.

Specifically, at least one site in a site sample test set is selected as a starting point, and the position information of the site and a site identifier for marking the integrity of the site information are extracted. It should be noted that, taking a blood sampling device as an example, in the blood sampling process, due to reasons such as the fluidity of blood, the thickness of blood of a user to be sampled, the penetration depth of a blood sampling needle, and the like, the blood coverage rate of each site on the test paper is different, and therefore, sites meeting the detection requirement standards need to be selected to form a site sample test set, that is, sites with complete site information are selected. And taking the position point as a starting point, selecting a certain number of position points according to row and column units in sequence to extract position information, and counting the number. Furthermore, according to the ratio of the number to the number of units of the acquisition length of the preset acquisition equipment, the first preset number is determined, and the corresponding position information is used as the principal component characteristics (namely, the first K principal component characteristics selected above).

It should be noted that the first preset number is at least 2, and at most does not exceed the maximum detection range that can be accommodated by the detection assembly of the acquisition device, and the specific number depends on the specific situation of the test feature set of the sample at the current site, and the embodiment of the present application is not particularly limited. However, in order to make the prediction result more accurate, in general, as many pieces of position information as possible within the maximum detection range are selected as principal component features.

In order to avoid the problem of low efficiency of manual prediction and improve the accuracy of a risk prediction result, in this embodiment of the application, further optionally, before performing risk prediction processing on medical detection data based on a risk prediction model that has been trained by a model, the method of this embodiment further includes: and constructing a convolutional neural network model, and performing model training on the convolutional neural network model based on the medical detection data sample set to obtain the risk prediction model.

The risk prediction model is used for configuring the hierarchical weight of the convolutional neural network model based on the number of the test site samples in the medical detection data sample set to complete iterative training.

Specifically, the embodiment of the present application selects to construct a convolutional neural network model, i.e., an initial prediction model. And training the historical medical detection data sample set to obtain a prediction model.

In order to make the referred user portrait data more comprehensive, preferably, the method of this embodiment further includes: acquiring user basic data matched with a target disease condition in a user medical database, and disease condition time, treatment time and a medical stage associated with the target disease condition; and establishing a target disease user image database based on the user basic data, the disease time, the treatment time and the medical stage.

In one possible implementation, the data is medical data, such as personal health records, prescriptions, exam reports, and the like. Specifically, a medical database of the medical institution may be traversed, users matched with the target medical condition may be screened out, and basic data of the users (for example, information data of user age, gender, residence, and the like) and medical condition time, treatment time, medical stage, and the like associated with the target medical condition may be acquired as reference data for information pushing. For example, the target user currently performing risk identification is a female with diabetic kidney symptoms in the age of 45, the physical condition is expressed as initial symptoms, and when information is pushed, the information is matched with the treatment strategy similar to the age of the female user and pushed with the treatment strategy of the female user in the initial symptoms so as to ensure the accuracy of information pushing. Further, a target disease user image database is established based on the user basic data, the disease time, the treatment time and the medical stage.

In the embodiment of the present application, further, after the target disorder user image database is established based on the user basic data, the disorder time, the treatment time, and the medical stage, the method of the present embodiment further includes: acquiring diagnosis information generated in the visiting process of each user in a target disease user picture database, current symptom information and expected symptom information matched with the diagnosis information; and generating medical risk information matched with different medical prediction results by combining the diagnosis information, the current symptom information and the expected symptom information.

Wherein the expected symptom information is used for representing the symptom information which can be reached by the user after the treatment in the database, for example, the user with the middle-stage symptom can be expected to recover to the early-stage symptom after the active coordination treatment. Specifically, after the target disorder user image database is established, diagnosis information generated by a total number of users in the diagnosis process in the database, current symptom information and expected symptom information matched with the diagnosis information are obtained, and further, medical risk information matched with different medical prediction results is generated, for example, the user A is in early stage of diabetes and renal symptoms, and based on a treatment strategy formulated by the current diagnosis information, the disorder can be expected to be relieved to be the same as a normal index, and only a little control is needed, so that the disorder risk state of the user A is a low-level risk. When the B user risk status for risk prediction is low-level risk, the B user may push the treatment strategy of the a user to the B user by further matching based on the diagnosis information, and the current symptom information and the expected symptom information matching with the diagnosis information, and concluding that the B user is matched with the a user.

In order to improve the accuracy of information pushing, in this embodiment of the application, it is further preferable that after the medical risk information matched with the medical prediction result in the user portrait data is obtained and pushed, the method of this embodiment further includes: recording the receiving times of the medical risk information, and collecting medical operations generated by each user based on the medical risk information; and updating the preset similarity threshold according to the similarity of the medical operation and the visit information of the first process information.

Specifically, after the information push is completed, the number of times the medical risk information is received and the medical operation performed by the user who receives the medical risk information may be recorded. The medical operation may be a treatment strategy taken, or diagnostic information after a visit, etc. Furthermore, the preset similarity threshold is updated to ensure the accuracy of the prediction result and the information push.

The application provides a medical risk information pushing method based on machine learning. Compared with the prior art, the method and the device have the advantages that all the site information of the preset acquisition equipment is obtained; screening target test sites from all the site information based on a screening processing model, and extracting medical detection data matched with the target test sites, wherein the screening processing model is obtained by adjusting model hyper-parameters based on model evaluation indexes of a test validation set and completing training; performing risk prediction processing on the medical detection data based on a risk prediction model which is trained by the model, and analyzing first process information matched with target disease information in a medical prediction result obtained after the risk prediction processing, wherein the first process information is used for representing disease time, treatment information and a medical stage matched with the target disease process information; if the similarity between second process information corresponding to each user portrait data in the established target disease user portrait database and the first process information exceeds a preset similarity threshold, medical risk information matched with the medical prediction result in the user portrait data is obtained and pushed, the information can be pushed timely and flexibly according to the self condition of a patient, the effectiveness of information pushing is improved, the incidence rate of the illness risk is reduced, and meanwhile, the accuracy of processing medical data in the intelligent medical system is improved.

Further, as an implementation of the method shown in fig. 1, an embodiment of the present application provides a medical risk information pushing device based on machine learning, as shown in fig. 2, the device includes: the system comprises a first obtaining module 21, a screening module 22, an analyzing module 23 and a pushing module 24.

The first acquisition module 21 is configured to acquire all site information of a preset acquisition device;

the screening module 22 is configured to screen a target test site from the information of all sites based on a screening processing model, and extract medical detection data matched with the target test site, where the screening processing model is obtained by adjusting a model hyper-parameter based on a model evaluation index of a test validation set and completing training;

the analysis module 23 is configured to perform risk prediction processing on the medical detection data based on a risk prediction model with model training completed, and analyze first process information matched with target disease information in a medical prediction result obtained after the risk prediction processing, where the first process information is used to represent disease time, treatment information, and a medical stage matched with the target disease course information;

and the pushing module 24 is configured to, if the similarity between the first process information and the second process information corresponding to each user portrait data in the established target disease user portrait database exceeds a preset similarity threshold, obtain medical risk information, which is matched with the medical prediction result, in the user portrait data, and push the medical risk information.

In a specific application scenario, before the first obtaining module 21, the apparatus further includes:

the determining module is used for acquiring a site sample test feature set and determining a first preset number of principal component features;

the calculation module is used for carrying out magnitude sequence sequencing through the coefficient absolute values of all the principal component characteristics, screening the site information of a second preset number, and calculating a model evaluation index based on the site sample test feature set;

and the training module is used for configuring and adjusting model hyperparameters through the model evaluation indexes and training the screening processing model by combining the screened site information and the model hyperparameters.

In a specific application scenario, the determining module includes:

the extraction unit is used for extracting at least one site position information in the site sample test set and a site mark for marking the integrity of the site information;

the statistical unit is used for selecting the site position information corresponding to at least two site identifications in the site sample test set according to row and column units in sequence and counting the number of the selected site position information;

and the determining unit is used for determining the position information corresponding to the first preset number as the principal component characteristic based on the ratio of the number to the number of the collection length units of the preset collection equipment.

In a specific application scenario, before the parsing module 23, the apparatus further includes:

the risk prediction model comprises a construction module and a risk prediction module, wherein the construction module is used for constructing a convolutional neural network model and carrying out model training on the convolutional neural network model based on a medical detection data sample set to obtain the risk prediction model, and the risk prediction model is used for completing iterative training by configuring the hierarchical weight of the convolutional neural network model based on the number of test site samples in the medical detection data sample set.

In a specific application scenario, the apparatus further includes:

the second acquisition module is used for acquiring user basic data matched with a target disease state in a user medical database, and disease state time, treatment time and a medical stage which are associated with the target disease state;

and the establishing module is used for establishing a target disease user image database based on the user basic data, the disease time, the treatment time and the medical stage.

In a specific application scenario, after the module is established, the apparatus further includes:

the third acquisition module is used for acquiring diagnosis information generated in the visiting process of each user in the target disease user image database, and current symptom information and expected symptom information matched with the diagnosis information;

and the generation module is used for combining the diagnosis information, the current symptom information and the expected symptom information to generate medical risk information matched with different medical prediction results.

In a specific application scenario, after the pushing module 24, the apparatus further includes:

the recording module is used for recording the receiving times of the medical risk information and acquiring medical operations generated by each user based on the medical risk information;

and the updating module is used for updating the preset similarity threshold according to the similarity between the medical operation and the visit information of the first process information.

The application provides a medical risk information pushing device based on machine learning. Compared with the prior art, the method and the device have the advantages that all the site information of the preset acquisition equipment is obtained; screening target test sites from all the site information based on a screening processing model, and extracting medical detection data matched with the target test sites, wherein the screening processing model is obtained by adjusting model hyper-parameters based on model evaluation indexes of a test validation set and completing training; performing risk prediction processing on the medical detection data based on a risk prediction model which is trained by the model, and analyzing first process information matched with target disease information in a medical prediction result obtained after the risk prediction processing, wherein the first process information is used for representing disease time, treatment information and a medical stage matched with the target disease process information; if the similarity between second process information corresponding to each user portrait data in the established target disease user portrait database and the first process information exceeds a preset similarity threshold, medical risk information matched with the medical prediction result in the user portrait data is obtained and pushed, the information can be pushed timely and flexibly according to the self condition of a patient, the effectiveness of information pushing is improved, the incidence rate of the illness risk is reduced, and meanwhile, the accuracy of processing medical data in the intelligent medical system is improved.

According to an embodiment of the present application, a storage medium is provided, where the storage medium stores at least one executable instruction, and the computer executable instruction may execute the medical risk information pushing method based on machine learning in any of the above method embodiments.

Based on such understanding, the technical solution of the present application may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (which may be a CD-ROM, a usb disk, a removable hard disk, etc.), and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the implementation scenarios of the present application.

Fig. 3 is a schematic structural diagram of a computer device according to an embodiment of the present application, and the specific embodiment of the present application does not limit a specific implementation of the computer device.

As shown in fig. 3, the computer apparatus may include: a processor (processor)302, a communication Interface 304, a memory 306, and a communication bus 308.

Wherein: the processor 302, communication interface 304, and memory 306 communicate with each other via a communication bus 308.

A communication interface 304 for communicating with network elements of other devices, such as clients or other servers.

The processor 302 is configured to execute the program 310, and may specifically execute relevant steps in the above data processing method embodiment based on project declaration.

In particular, program 310 may include program code comprising computer operating instructions.

The processor 302 may be a central processing unit CPU, or an application Specific Integrated circuit asic, or one or more Integrated circuits configured to implement embodiments of the present application. The computer device includes one or more processors, which may be the same type of processor, such as one or more CPUs; or may be different types of processors such as one or more CPUs and one or more ASICs.

And a memory 306 for storing a program 310. Memory 306 may comprise high-speed RAM memory and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.

The program 310 may specifically be configured to cause the processor 302 to perform the following operations:

acquiring all site information of preset acquisition equipment;

screening target test sites from all the site information based on a screening processing model, and extracting medical detection data matched with the target test sites, wherein the screening processing model is obtained by adjusting model hyper-parameters based on model evaluation indexes of a test validation set and completing training;

performing risk prediction processing on the medical detection data based on a risk prediction model which is trained by the model, and analyzing first process information matched with target disease information in a medical prediction result obtained after the risk prediction processing, wherein the first process information is used for representing disease time, treatment information and a medical stage matched with the target disease process information;

and if the similarity between the second process information corresponding to each user portrait data in the established target disease user portrait database and the first process information exceeds a preset similarity threshold, acquiring medical risk information matched with the medical prediction result in the user portrait data, and pushing the medical risk information.

The storage medium may further include an operating system and a network communication module. The operating system is a program for managing hardware and software resources of the entity device for processing the business data based on the multi-modal hybrid model, and supports the operation of an information processing program and other software and/or programs. The network communication module is used for realizing communication among components in the storage medium and communication with other hardware and software in the information processing entity device.

It will be apparent to those skilled in the art that the modules or steps of the present application described above may be implemented by a general purpose computing device, they may be centralized on a single computing device or distributed across a network of multiple computing devices, and alternatively, they may be implemented by program code executable by a computing device, such that they may be stored in a storage device and executed by a computing device, and in some cases, the steps shown or described may be performed in an order different than that described herein, or they may be separately fabricated into individual integrated circuit modules, or multiple ones of them may be fabricated into a single integrated circuit module. Thus, the present application is not limited to any specific combination of hardware and software.

The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.