Single disease health management knowledge graph generation method and device based on big data

Abstract

The present invention relates to a method and device for generating a single disease health management knowledge graph based on big data, the method comprising: collecting basic data of chronic disease population; preprocessing the basic data; learning relevant data of single disease chronic disease complications based on the preprocessed basic data, constructing a single disease chronic disease complication incidence prediction model and a single disease chronic disease complication severity prediction model; applying the initial health management knowledge graph to the single disease chronic disease complication incidence prediction model and the single disease chronic disease complication severity prediction model for enhanced deep learning to adjust the application weight of the initial health management knowledge graph to obtain a final health management knowledge graph; verifying the final health management knowledge graph, and saving it after verification. The present invention is adapted to the course of disease of different populations or individuals, and effectively supports the health management of single diseases.

Description

Technical Field

[0001] The present invention relates to the field of big data application technology, and in particular to a method and device for generating a single disease health management knowledge graph based on big data. Background Art

[0002] As people's living standards continue to improve, their attention to medical health is increasing year by year. How to mine medical information contained in medical records and diagnosis results has attracted more and more attention. Using text mining technology to process this knowledge into structured information can bring great progress to the informatization of medical knowledge. The rapid development of natural language processing makes it possible to automatically extract medical entities and relationships between entities from medical records. The extracted medical knowledge can be used to construct a medical knowledge graph to promote the development of medical intelligence.

[0003] The existing knowledge graph for single disease health management in chronic disease populations is manually compiled by summarizing expert experience. Therefore, in actual application, it cannot adapt well to the disease course of different populations or individuals, resulting in the inability to effectively support single disease health management. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method and device for generating a single disease health management knowledge graph based on big data, which is adaptable to the disease course of different populations or individuals and effectively supports the health management of a single disease.

[0005] The technical solution adopted by the present invention to solve the technical problem is: to provide a method for generating a single disease health management knowledge graph based on big data, comprising the following steps:

[0006] Collect basic data of chronic disease population;

[0007] Preprocessing the basic data;

[0008] Based on the pre-processed basic data, the relevant data of single-disease chronic disease complications are learned to construct a single-disease chronic disease complication incidence prediction model and a single-disease chronic disease complication severity prediction model;

[0009] Applying the initial health management knowledge graph to the single-disease chronic disease complication incidence prediction model and the single-disease chronic disease complication severity prediction model to perform enhanced deep learning to adjust the application weight of the initial health management knowledge graph to obtain a final health management knowledge graph;

[0010] The final health management knowledge graph is verified and saved after passing the verification.

[0011] The basic data of the chronic disease population includes physical examination data, medical treatment data and living habit data of the chronic disease population.

[0012] The preprocessing of the basic data comprises:

[0013] Performing standardized conversion on the basic data;

[0014] The basic data after standardization conversion is desensitized.

[0015] The learning of the relevant data of the single-disease chronic disease complications based on the pre-processed basic data to construct a single-disease chronic disease complication incidence prediction model and a single-disease chronic disease complication severity prediction model specifically includes:

[0016] Finding first data related to the incidence of complications of a single disease type chronic disease from the preprocessed basic data, and inputting the first data into a multi-dimensional distribution prediction model for learning to obtain a single disease type chronic disease complication incidence prediction model;

[0017] Find the second data related to the treatment of complications of a single chronic disease from the preprocessed basic data, and input the second data into the multidimensional distribution prediction model for learning to obtain a single disease chronic disease complication severity prediction model.

[0018] The initial health management knowledge graph is applied to the single-disease chronic disease complication incidence prediction model and the single-disease chronic disease complication severity prediction model for enhanced deep learning to adjust the application weight of the initial health management knowledge graph, and the final health management knowledge graph is obtained as follows:

[0019] Obtain the initial health management knowledge graph;

[0020] Performing enhanced deep learning on the single-disease chronic disease complication incidence prediction model, applying the initial health management knowledge graph to the single-disease chronic disease complication incidence prediction model for reference training, and adjusting the application weight of the initial health management knowledge graph according to the training results to obtain a mid-term health management knowledge graph;

[0021] The single-disease chronic disease complication severity prediction model is subjected to enhanced deep learning, and the mid-term health management knowledge graph is applied to the single-disease chronic disease complication severity prediction model for reference training, and the application weight of the mid-term health management knowledge graph is adjusted according to the training results to obtain the final health management knowledge graph.

[0022] The technical solution adopted by the present invention to solve the technical problem is: a device for generating a single disease health management knowledge graph based on big data is also provided, comprising:

[0023] Collection module, used to collect basic data of chronic disease population;

[0024] A preprocessing module, used for preprocessing the basic data;

[0025] A model building module, used to learn the relevant data of single-disease chronic disease complications based on the preprocessed basic data, and to build a single-disease chronic disease complication incidence prediction model and a single-disease chronic disease complication severity prediction model;

[0026] A knowledge graph generation module, used for applying the initial health management knowledge graph to the single-disease chronic disease complication incidence prediction model and the single-disease chronic disease complication severity prediction model to perform enhanced deep learning to adjust the application weight of the initial health management knowledge graph to obtain a final health management knowledge graph;

[0027] The knowledge graph verification module is used to verify the final health management knowledge graph and save it after the verification is passed.

[0028] The basic data of the chronic disease population includes physical examination data, medical treatment data and living habit data of the chronic disease population.

[0029] The preprocessing module comprises:

[0030] A standardization conversion unit, used for performing standardization conversion on the basic data;

[0031] The desensitizing unit is used to perform desensitizing processing on the basic data after the standardization conversion.

[0032] The model building module includes:

[0033] A first model building unit is used to find out first data related to the incidence of complications of a single disease type chronic disease from the preprocessed basic data, and input the first data into a multi-dimensional distribution prediction model for learning to obtain a single disease type chronic disease complication incidence prediction model;

[0034] The second model building unit is used to find out the second data related to the treatment of complications of a single disease from the preprocessed basic data, and input the second data into the multi-dimensional distribution prediction model for learning to obtain a single disease chronic disease complication severity prediction model.

[0035] The knowledge graph generation module includes:

[0036] An acquisition unit, used to acquire an initial health management knowledge graph;

[0037] The first learning unit is used to perform enhanced deep learning on the single-disease chronic disease complication incidence prediction model, and apply the initial health management knowledge graph to the single-disease chronic disease complication incidence prediction model for reference training, and adjust the application weight of the initial health management knowledge graph according to the training result to obtain a mid-term health management knowledge graph;

[0038] The second learning unit is used to perform enhanced deep learning on the single disease chronic disease complication severity prediction model, and apply the mid-term health management knowledge graph to the single disease chronic disease complication severity prediction model for reference training, and adjust the application weight of the mid-term health management knowledge graph according to the training results to obtain the final health management knowledge graph.

[0039] Beneficial Effects

[0040] Due to the adoption of the above-mentioned technical scheme, compared with the prior art, the present invention has the following advantages and positive effects: the present invention is based on the collection of big data information such as physical examinations, medical treatments, and living habits of chronic disease populations, and studies the correlation between health management and chronic disease complications. On the basis of the traditional chronic disease health management knowledge graph, the present invention further strengthens the training of a knowledge graph that conforms to the health management of a single disease. The obtained knowledge graph adapts to the course of disease of different populations or individuals, and effectively supports the health management of a single disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a flowchart of the first embodiment. DETAILED DESCRIPTION

[0042] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.

[0043] The first embodiment of the present invention relates to a method for generating a single disease health management knowledge graph based on big data, comprising the following steps:

[0044] Step 1, collect the basic data of the chronic disease population and pre-process the basic data. In this step, the basic data of the chronic disease population collected includes the physical examination data, medical data and living habit data of the chronic disease population. Among them, the pre-processing of the basic data includes: step 1a, standardizing the basic data, through which a variety of heterogeneous data sources can be uniformly processed to improve data processing efficiency; step 1b, desensitizing the basic data after standardization conversion, through which sensitive privacy data can be reliably protected.

[0045] Step 2, machine learning of single-disease chronic disease complications, that is, learning the relevant data of single-disease chronic disease complications based on the pre-processed basic data, constructing a single-disease chronic disease complication incidence prediction model and a single-disease chronic disease complication severity prediction model, specifically including: Step 2a, finding the first data related to the incidence of single-disease chronic disease complications from the pre-processed basic data, and inputting the first data into the multi-dimensional distribution prediction model for learning, to obtain a single-disease chronic disease complication incidence prediction model; Step 2b, finding the second data related to the treatment of single-disease chronic disease complications from the pre-processed basic data, and inputting the second data into the multi-dimensional distribution prediction model for learning, to obtain a single-disease chronic disease complication severity prediction model. This implementation method studies the correlation between health management and chronic disease complications based on big data information such as the incidence and severity of complications in chronic disease populations, laying a foundation for subsequent reinforcement learning.

[0046] Step 3, complete the enhanced deep learning migration based on the initial health management knowledge graph, that is, apply the initial health management knowledge graph to the single disease chronic disease complication incidence prediction model and the single disease chronic disease complication severity prediction model for enhanced deep learning to adjust the application weight of the initial health management knowledge graph to obtain the final health management knowledge graph, specifically including: Step 3a, obtain the initial health management knowledge graph, and use the basic data of the above-mentioned chronic disease population as the training source data, wherein the initial health management knowledge graph in this embodiment can be a health management knowledge graph summarized by traditional expert experience, or it can be a health management knowledge graph obtained using this embodiment. Step 3b, strengthen deep learning of the single disease chronic disease complication incidence prediction model, and apply the initial health management knowledge graph to the single disease chronic disease complication incidence prediction model for reference training, and adjust the application weight of the initial health management knowledge graph according to the training results to obtain the mid-term health management knowledge graph; Step 3c, strengthen deep learning of the single disease chronic disease complication severity prediction model, and apply the mid-term health management knowledge graph to the single disease chronic disease complication severity prediction model for reference training, and adjust the application weight of the mid-term health management knowledge graph according to the training results to obtain the final health management knowledge graph. Based on the traditional chronic disease health management knowledge graph, this implementation further strengthens the training of a knowledge graph that conforms to single disease health management, improves the training speed through transfer learning, and ensures the accuracy of the obtained single disease health management knowledge graph in adapting to personal characteristics.

[0047] Step 4, generating a single disease health management knowledge graph, specifically includes: step 4a, forming a strategy model based on the final health management knowledge graph obtained by the enhanced deep learning training in step 3; step 4b, using the basic data of the above-mentioned chronic disease population as the simulation verification source data, verifying the accuracy of the strategy model in adapting to personal characteristics. When the verification passes, the final health management knowledge graph can be saved. If the verification fails, return to step 3 and re-perform enhanced deep learning migration based on the health management knowledge graph.

[0048] The second embodiment of the present invention relates to a device for generating a single disease health management knowledge graph based on big data, comprising: a collection module for collecting basic data of a chronic disease population; a preprocessing module for preprocessing the basic data; a model building module for learning relevant data of single disease chronic disease complications based on the preprocessed basic data, and constructing a single disease chronic disease complication incidence prediction model and a single disease chronic disease complication severity prediction model; a knowledge graph generation module for applying the initial health management knowledge graph to the single disease chronic disease complication incidence prediction model and the single disease chronic disease complication severity prediction model for enhanced deep learning to adjust the application weight of the initial health management knowledge graph to obtain the final health management knowledge graph; a knowledge graph verification module for verifying the final health management knowledge graph and saving it after the verification is passed.

[0049] The basic data of the chronic disease population includes physical examination data, medical treatment data and living habit data of the chronic disease population.

[0050] The preprocessing module includes: a standardization conversion unit, which is used to perform standardization conversion on the basic data; and a desensitization unit, which is used to perform desensitization processing on the basic data after the standardization conversion.

[0051] The model building module includes: a first model building unit, which is used to find out the first data related to the incidence of complications of a single disease type of chronic disease from the preprocessed basic data, and input the first data into the multidimensional distribution prediction model for learning to obtain a single disease type of chronic disease complication incidence prediction model; a second model building unit, which is used to find out the second data related to the treatment of complications of a single disease type of chronic disease from the preprocessed basic data, and input the second data into the multidimensional distribution prediction model for learning to obtain a single disease type of chronic disease complication severity prediction model.

[0052] The knowledge graph generation module includes: an acquisition unit, used to acquire an initial health management knowledge graph; a first learning unit, used to perform enhanced deep learning on the single disease chronic disease complication incidence prediction model, and apply the initial health management knowledge graph to the single disease chronic disease complication incidence prediction model for reference training, and adjust the application weight of the initial health management knowledge graph according to the training results to obtain a mid-term health management knowledge graph; a second learning unit, used to perform enhanced deep learning on the single disease chronic disease complication severity prediction model, and apply the mid-term health management knowledge graph to the single disease chronic disease complication severity prediction model for reference training, and adjust the application weight of the mid-term health management knowledge graph according to the training results to obtain a final health management knowledge graph.

[0053] It is not difficult to find that the present invention is based on the collection of big data information such as physical examinations, medical treatments, and living habits of people with chronic diseases, to study the correlation between health management and chronic disease complications, and on the basis of the traditional chronic disease health management knowledge graph, further strengthens the training of a knowledge graph that conforms to the health management of a single disease. The obtained knowledge graph is adapted to the course of disease of different populations or individuals, and effectively supports the health management of a single disease.

Claims

1. A method for generating a single disease health management knowledge graph based on big data, characterized in that: The following steps are involved: Collect basic data of chronic disease population; Preprocessing the basic data; Based on the pre-processed basic data, the relevant data of single-disease chronic disease complications are learned to construct a single-disease chronic disease complication incidence prediction model and a single-disease chronic disease complication severity prediction model; The initial health management knowledge graph is applied to the single-disease chronic disease complication incidence prediction model and the single-disease chronic disease complication severity prediction model for enhanced deep learning to adjust the application weight of the initial health management knowledge graph to obtain the final health management knowledge graph, which is specifically: Obtain the initial health management knowledge graph; Performing enhanced deep learning on the single-disease chronic disease complication incidence prediction model, applying the initial health management knowledge graph to the single-disease chronic disease complication incidence prediction model for reference training, and adjusting the application weight of the initial health management knowledge graph according to the training results to obtain a mid-term health management knowledge graph; Performing enhanced deep learning on the single-disease chronic disease complication severity prediction model, applying the mid-term health management knowledge graph to the single-disease chronic disease complication severity prediction model for reference training, and adjusting the application weight of the mid-term health management knowledge graph according to the training results to obtain the final health management knowledge graph; The final health management knowledge graph is verified and saved after passing the verification.

2. The method for generating a single disease health management knowledge graph based on big data according to claim 1 is characterized in that: The basic data of the chronic disease population includes physical examination data, medical treatment data and living habit data of the chronic disease population.

3. The method for generating a single disease health management knowledge graph based on big data according to claim 1 is characterized in that: The preprocessing of the basic data comprises: Performing standardized conversion on the basic data; The basic data after standardization conversion is desensitized.

4. The method for generating a single disease health management knowledge graph based on big data according to claim 1 is characterized in that: The learning of the relevant data of the single-disease chronic disease complications based on the pre-processed basic data to construct a single-disease chronic disease complication incidence prediction model and a single-disease chronic disease complication severity prediction model specifically includes: Finding first data related to the incidence of complications of a single disease type chronic disease from the preprocessed basic data, and inputting the first data into a multi-dimensional distribution prediction model for learning to obtain a single disease type chronic disease complication incidence prediction model; Find the second data related to the treatment of complications of a single chronic disease from the preprocessed basic data, and input the second data into the multidimensional distribution prediction model for learning to obtain a single disease chronic disease complication severity prediction model.

5. A device for generating a single disease health management knowledge graph based on big data, characterized in that: include: Collection module, used to collect basic data of chronic disease population; A preprocessing module, used for preprocessing the basic data; A model building module, used to learn the relevant data of single-disease chronic disease complications based on the preprocessed basic data, and to build a single-disease chronic disease complication incidence prediction model and a single-disease chronic disease complication severity prediction model; A knowledge graph generation module, used for applying the initial health management knowledge graph to the single-disease chronic disease complication incidence prediction model and the single-disease chronic disease complication severity prediction model to perform enhanced deep learning to adjust the application weight of the initial health management knowledge graph to obtain a final health management knowledge graph; The knowledge graph generation module includes: An acquisition unit, used to acquire an initial health management knowledge graph; The first learning unit is used to perform enhanced deep learning on the single-disease chronic disease complication incidence prediction model, and apply the initial health management knowledge graph to the single-disease chronic disease complication incidence prediction model for reference training, and adjust the application weight of the initial health management knowledge graph according to the training result to obtain a mid-term health management knowledge graph; The second learning unit is used to perform enhanced deep learning on the single-disease chronic disease complication severity prediction model, and apply the mid-term health management knowledge graph to the single-disease chronic disease complication severity prediction model for reference training, and adjust the application weight of the mid-term health management knowledge graph according to the training result to obtain the final health management knowledge graph; The knowledge graph verification module is used to verify the final health management knowledge graph and save it after the verification is passed.

6. The device for generating a single disease health management knowledge graph based on big data according to claim 5 is characterized in that: The basic data of the chronic disease population includes physical examination data, medical treatment data and living habit data of the chronic disease population.

7. The device for generating a single disease health management knowledge graph based on big data according to claim 5 is characterized in that: The preprocessing module comprises: A standardization conversion unit, used for performing standardization conversion on the basic data; The desensitizing unit is used to perform desensitizing processing on the basic data after the standardization conversion.

8. The device for generating a single disease health management knowledge graph based on big data according to claim 5 is characterized in that: The model building module includes: A first model building unit is used to find out first data related to the incidence of complications of a single disease type chronic disease from the preprocessed basic data, and input the first data into a multi-dimensional distribution prediction model for learning to obtain a single disease type chronic disease complication incidence prediction model; The second model building unit is used to find out the second data related to the treatment of complications of a single disease from the preprocessed basic data, and input the second data into the multi-dimensional distribution prediction model for learning to obtain a single disease chronic disease complication severity prediction model.

Images