A diabetes risk prediction system based on multiple models
By constructing a multi-model diabetes risk prediction system, and utilizing deep learning and neural network models, feature extraction and risk prediction of patient information are achieved, solving the problem that existing technologies cannot effectively assess diabetes risk and improving the accuracy and comprehensiveness of the assessment.
Patent Information
- Application Number
- CN202311757132.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-19
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2043-12-19
AI Technical Summary
Existing technologies lack diabetes risk assessment systems based on the cooperation of multiple models, making it impossible to effectively predict whether a patient has diabetes and its type.
A multi-model-based diabetes risk prediction system is constructed, including a learning model construction module, an attribute information identification and extraction module, a directed graph model construction module, a node information identification and extraction module, a feature vector extraction module, and a network model construction module. By combining these modules, a deep learning model for diabetes and a neural network model for diabetes types are constructed to achieve feature extraction and risk prediction of patient information.
It can accurately predict whether a patient has a certain type of diabetes and the associated risk, thus improving the accuracy and comprehensiveness of diabetes risk assessment.
Smart Images

Figure CN117747121B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of diabetes risk prediction, and in particular to a diabetes risk prediction system based on multiple models. Background Art
[0002] Diabetes is a common endocrine metabolic disease caused by high blood sugar due to relative or absolute lack of insulin secretion, which leads to protein and fat metabolism disorders. The symptoms of diabetes are often manifested as polydipsia, polyphagia, polyuria, and weight loss, but the main harm is that long-term high blood sugar in patients will cause various serious complications, causing great harm to organs such as kidneys, blood vessels, and nerves, and even endangering life.
[0003] Currently, there are diabetes assessment technologies based on machine learning that are used to assess whether a patient has diabetes, etc. However, there is no system that cooperates with multiple models (diabetes deep learning model, diabetes type directed graph model corresponding to different diabetes types, and diabetes type neural network model) to achieve diabetes risk assessment. Summary of the invention
[0004] In view of the problems and shortcomings of the prior art, the present invention provides a diabetes risk prediction system based on multiple models.
[0005] The present invention solves the above technical problems through the following technical solutions:
[0006] The present invention provides a diabetes risk prediction system based on multiple models, which is characterized in that it includes a learning model construction module, an attribute information recognition and extraction module, a directed graph model construction module, a node information recognition and extraction module, a feature vector extraction module and a network model construction module.
[0007] The learning model building module is used to train a deep learning model based on the obtained diabetes-related information to build a diabetes deep learning model.
[0008] The attribute information identification and extraction module is used to preconfigure attribute information identification and extraction rules. The diabetes deep learning model uses the attribute information identification and extraction rules to identify and extract multiple historical information of diabetes patients, from which multiple basic information attributes and multiple information attribute directional relationships under different diabetes types are identified and extracted. The information attribute directional relationship is a directional relationship of the first information attribute → the second information attribute.
[0009] The directed graph model construction module is used to construct a directed graph model of the diabetes type under the corresponding diabetes type based on multiple basic information attributes and multiple information attribute directional relationships under different diabetes types and graph theory principles, and the basic information attribute nodes and attribute nodes that constitute the information attribute directional relationships in the directed graph model of the diabetes type reach the target number as priority attribute nodes. When the priority attribute node is the second information attribute node, its corresponding first information attribute node is also a priority attribute node, and the rest are non-priority attribute nodes.
[0010] The node information identification and extraction module is used to identify and extract historical information of multiple diabetes patients of the corresponding diabetes type based on the pre-configured node information identification and extraction rules corresponding to each diabetes type directed graph model. The diabetes deep learning model uses the node information identification and extraction rules to identify and extract historical information of multiple diabetes patients of the corresponding diabetes type. The specific content of the priority attribute nodes under the diabetes type directed graph model is first extracted from the historical information of each diabetes patient and filled into the corresponding priority attribute nodes, and then the specific content of the non-priority attribute nodes is extracted and filled into the corresponding non-priority attribute nodes, thereby obtaining a diabetes type directed graph for each diabetes patient under each diabetes type.
[0011] The feature vector extraction module is used to extract a directed feature vector in which both the first information attribute node and the second information attribute node are filled from the diabetes type directed graph of each diabetes patient under each diabetes type.
[0012] The network model building module is used to take all directional feature vectors extracted from each diabetic patient under each diabetes type as an input sample set and the corresponding diabetes type as an output sample to train the convolutional neural network model to obtain a diabetes type neural network model.
[0013] The diabetes risk prediction system also includes a diabetes risk prediction module, which is used to receive information of a patient to be estimated, obtain each diabetes type directed graph model, call the node information identification and extraction module for each diabetes type directed graph model, and the diabetes deep learning model uses the node information identification and extraction rules corresponding to the diabetes type directed graph model to identify and extract the information of the patient to be estimated, first extract the specific content of the priority attribute node under the diabetes type directed graph model from the information of the patient to be estimated, fill it into the corresponding priority attribute node, and then extract the specific content of the non-priority attribute node and fill it into the corresponding non-priority attribute node, so as to obtain each diabetes type of the patient to be estimated. A diabetes type directed graph is obtained, and then a feature vector extraction module is called to extract a directed feature vector in which the first information attribute node and the second information attribute node are both filled from each diabetes type directed graph. The directed feature vector extracted from the diabetes type directed graph is input into the diabetes type neural network model as a separate input sample set to output whether the patient to be estimated has the risk of the diabetes type corresponding to the diabetes type directed graph. If the answer is yes, risk information of the patient to be estimated having the diabetes type is issued; if the answer is no, risk information of the patient to be estimated not having the diabetes type is issued, thereby estimating whether the patient to be estimated has the risk of diabetes and the risk of which type of diabetes he has.
[0014] The positive and progressive effects of the present invention are:
[0015] The present invention constructs a diabetes deep learning model, a diabetes type directed graph model corresponding to different diabetes types, and a diabetes type neural network model. By combining these multiple models, the risk of a patient to be estimated suffering from a certain type of diabetes can be predicted. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 4 is a structural block diagram of a diabetes risk prediction system according to a preferred embodiment of the present invention.
[0017] Figure 2 This is a directed graph model diagram of diabetes types according to a preferred embodiment of the present invention.
[0018] Figure 3 A schematic diagram of constructing a directed graph model according to a preferred embodiment of the present invention. DETAILED DESCRIPTION
[0019] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0020] like Figure 1 As shown, this embodiment provides a diabetes risk prediction system based on multiple models, which includes a learning model construction module 1, an attribute information identification and extraction module 2, a directed graph model construction module 3, a node information identification and extraction module 4, a feature vector extraction module 5, a network model construction module 6 and a diabetes risk prediction module 7.
[0021] The learning model construction module 1 is used to train a deep learning model based on the obtained diabetes-related information to construct a diabetes deep learning model.
[0022] Diabetes-related information includes data in the medical field of diabetes, diet data, exercise data and sleep data corresponding to diabetes, medical data on diabetes in various hospitals and research data related to diabetic foot, symptom data, treatment drug data, diet data, exercise data and sleep data of diabetic patients, and diabetes information crawled from the Internet using crawler tools.
[0023] The attribute information identification and extraction module 2 is used to preconfigure attribute information identification and extraction rules. The diabetes deep learning model uses the attribute information identification and extraction rules to identify and extract historical information of multiple diabetes patients, from which multiple basic information attributes and multiple information attribute directional relationships under different diabetes types are identified and extracted. The information attribute directional relationship is a directional relationship of the first information attribute → the second information attribute.
[0024] The historical information of diabetic patients includes the type of diabetes suffered by diabetic patients, basic information of diabetic patients, symptom information of diabetic patients, examination information of diabetic patients and life information of diabetic patients. The basic information of diabetic patients includes the gender, age, weight of diabetic patients, and the life information of diabetic patients includes the diet information, exercise information and sleep information of diabetic patients.
[0025] The basic information attributes include the gender, age, and weight of the diabetic patient.
[0026] The directed graph model construction module 3 is used to construct a directed graph model of the diabetes type under the corresponding diabetes type based on multiple basic information attributes and multiple information attribute directional relationships under different diabetes types and graph theory principles, and the basic information attribute nodes and attribute nodes that constitute the information attribute directional relationship in the directed graph model of the diabetes type that reach the target number are used as priority attribute nodes. When the priority attribute node is the second information attribute node, its corresponding first information attribute node is also a priority attribute node, and the rest are used as non-priority attribute nodes.
[0027] Basic information attribute nodes such as gender attribute node, age attribute node and weight attribute node are used as priority attribute nodes. Figure 2 It is a directed graph model of diabetes type, in which gender attribute node A2, age attribute node A3 and weight attribute node A4 are priority attribute nodes. In the directed graph model of diabetes type, the number of attribute node A1 constituting information attribute directional relationships is 7, which reaches the target number, the number of attribute node A5 constituting information attribute directional relationships is 3, which reaches the target number, the number of attribute node A11 constituting information attribute directional relationships is 3, which reaches the target number, and the number of attribute node A7 constituting information attribute directional relationships is 3, which reaches the target number. Attribute node A1, attribute node A5, attribute node A7 and attribute node A11 are priority attribute nodes. When priority attribute node A11 is the second information attribute node, its corresponding first information attribute node A6 is also a priority attribute node, and the rest are non-priority attribute nodes.
[0028] Directed graph model construction module 3 The process of constructing a directed graph model is as follows: Figure 3 Suppose there are n diabetes types, each diabetes type corresponds to a diabetes type directed graph model, and the nodes of the diabetes type directed graph model are divided into priority attribute nodes and non-priority attribute nodes.
[0029] The node information identification and extraction module 4 is used to identify and extract historical information of multiple diabetic patients of the corresponding diabetes type based on the pre-configured node information identification and extraction rules corresponding to each diabetes type directed graph model. The diabetes deep learning model uses the node information identification and extraction rules to identify and extract historical information of multiple diabetic patients of the corresponding diabetes type. The specific content of the priority attribute nodes under the diabetic type directed graph model is first extracted from the historical information of each diabetic patient and filled into the corresponding priority attribute nodes, and then the specific content of the non-priority attribute nodes is extracted and filled into the corresponding non-priority attribute nodes, thereby obtaining a diabetes type directed graph for each diabetic patient under each diabetes type.
[0030] Among them, the node information identification and extraction module 4 is used to first extract the specific content of the priority attribute node under the directed graph model of the diabetes type from the historical information of each diabetic patient and directly fill it into the corresponding priority attribute node; then extract the specific content of the non-priority attribute node and fill it into the corresponding non-priority attribute node: first extract the specific content of the non-priority attribute node that forms a directed relationship with the priority attribute node and fill it into the corresponding non-priority attribute node, and then extract the specific content of the non-priority attribute node that forms a directed relationship with the non-priority attribute node and fill it into the corresponding non-priority attribute node.
[0031] In this embodiment, the specific content of the extracted priority attribute node is directly filled into the corresponding priority attribute node. The specific content of the extracted non-priority attribute node is not directly filled into the corresponding non-priority attribute node, but needs to be processed according to the situation. The processing order is to first process the non-priority attribute node that has a directed relationship with the priority attribute node, and then process the non-priority attribute node that has a directed relationship with the non-priority attribute node.
[0032] The node information identification and extraction module 4 is used for a non-priority attribute node. The specific content of the non-priority attribute node has been filled. At present, new specific content of the non-priority attribute node is extracted. There is a situation where the new specific content and the filled specific content cannot be identified as consistent. In this case, this non-priority attribute node is an uncertain node, and the specific content of the non-priority attribute node is manually corrected and the non-priority attribute node is not marked as an uncertain node.
[0033] The feature vector extraction module 5 is used to extract a directed feature vector in which both the first information attribute node and the second information attribute node are filled from the diabetes type directed graph of each diabetes patient under each diabetes type. Since the uncertain nodes have been manually corrected, the directed feature vectors obtained here are all determined directed feature vectors.
[0034] The network model building module 6 is used to train the convolutional neural network model by taking all the directional feature vectors extracted from each diabetic patient under each diabetes type as an input sample set and the corresponding diabetes type as an output sample to obtain a diabetes type neural network model. The diabetes type neural network model can be used to predict whether the patient to be estimated has a certain type of diabetes risk.
[0035] The diabetes risk estimation module 7 is used to receive the information of a patient to be estimated, obtain each diabetes type directed graph model, call the node information identification and extraction module for each diabetes type directed graph model, and the diabetes deep learning model uses the node information identification and extraction rules corresponding to the diabetes type directed graph model to identify and extract the information of the patient to be estimated, first extract the specific content of the priority attribute node under the diabetes type directed graph model from the information of the patient to be estimated, fill it into the corresponding priority attribute node, then extract the specific content of the non-priority attribute node, fill it into the corresponding non-priority attribute node and mark the uncertain node, so as to obtain the diabetes type directed graph model under each diabetes type of the patient to be estimated. Figure, and then call the feature vector extraction module to extract the first information attribute node and the second information attribute node from each diabetes type directed graph as a directed feature vector, and input the directed feature vector extracted from the diabetes type directed graph as a separate input sample set into the diabetes type neural network model to output whether the patient to be estimated has the risk of the diabetes type corresponding to the diabetes type directed graph. If the answer is yes, the risk information of the patient to be estimated having the diabetes type is issued, and if the answer is no, the risk information of the patient to be estimated not having the diabetes type is issued, so as to estimate whether the patient to be estimated has the risk of diabetes and the risk of which type of diabetes he has.
[0036] Further, the diabetes risk estimation module 7 is used to first extract the specific content of the priority attribute node under the diabetes type directed graph model from the information of the patient to be estimated, fill it into the corresponding priority attribute node, and then extract the specific content of the non-priority attribute node and fill it into the corresponding non-priority attribute node, and mark it as an uncertain node when the filled specific content is uncertain, so as to obtain the diabetes type directed graph under each diabetes type of the patient to be estimated, and then call the feature vector extraction module to extract from each diabetes type directed graph the first information attribute node and the second information attribute node, both of which are filled and do not contain marked uncertain nodes as a determined directed feature vector, and the uncertain nodes that are filled and contain marks as a to-be-determined directed feature vector. When the number of determined directed feature vectors extracted from at least one diabetes type directed graph reaches a set number, the determined directed feature vectors extracted from the diabetes type directed graph will be preferentially input as a separate input sample set into the diabetes type neural network model to output whether the patient to be estimated has the risk of the diabetes type corresponding to the diabetes type directed graph. If it is yes, the information that the patient to be estimated has the risk of the diabetes type and the diabetes risk estimation is completed is issued;
[0037] The diabetes risk estimation module 7 is also used to issue a risk that the patient to be estimated does not have the type of diabetes when the answer is no, and display the to-be-determined directional feature vectors corresponding to the diabetes type directed graph that has not reached the set number and their original sources for manual correction so as to correct the to-be-determined directional feature vectors to determined directional feature vectors. When it is determined that the number of determined directional feature vectors corresponding to a corrected diabetes type directed graph reaches the set number, the determined directional feature vectors extracted from the corrected diabetes type directed graph are input as a separate input sample set into the diabetes type neural network model to output whether the patient to be estimated has the risk of the diabetes type corresponding to the corrected diabetes type directed graph. When the answer is yes, the risk of the patient to be estimated having the diabetes type and the information that the diabetes risk estimation is completed are issued. When the answer is no, the risk of the patient to be estimated not having the diabetes type is issued, and it is determined whether there are still to-be-determined directional feature vectors corresponding to the diabetes type directed graph that have not reached the set number and need to be corrected. When the answer is yes, wait for manual correction to continue. When the answer is no, the risk of the patient to be estimated not having any diabetes type and the information that the diabetes risk estimation is completed are issued.
[0038] The diabetes risk estimation module 7 is used to call the node information identification and extraction module 4 for each diabetes type directed graph model, first extract the specific content of the priority attribute node under the diabetes type directed graph model from the information of the patient to be estimated, and directly fill it into the corresponding priority attribute node; then extract the specific content of the non-priority attribute node and fill it into the corresponding non-priority attribute node, and mark it as an uncertain node when the specific content filled is uncertain, specifically, first extract the specific content of the non-priority attribute node that forms a directed relationship with the priority attribute node and fill it into the corresponding non-priority attribute node, and mark it as an uncertain node when the specific content filled is uncertain, then extract the specific content of the non-priority attribute node that forms a directed relationship with the non-priority attribute node and fill it into the corresponding non-priority attribute node, and mark it as an uncertain node when the specific content filled is uncertain. Among them, when the specific content filled is uncertain, it is marked as an uncertain node: the specific content of a non-priority attribute node has been filled, and the new specific content of the non-priority attribute node is extracted at present, and the new specific content and the filled specific content cannot be identified as the same content, then this non-priority attribute node is an uncertain node and the source of the new specific content and the source of the filled specific content are recorded for manual correction.
[0039] The information of the patient to be estimated includes the basic information of the patient to be estimated, the symptom information of the patient to be estimated, the examination information of the patient to be estimated and the living information of the patient to be estimated. The basic information of the patient to be estimated includes the gender, age and weight of the patient to be estimated, and the living information of the patient to be estimated includes the diet information, exercise information and sleep information of the patient to be estimated.
[0040] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that these are only examples, and the protection scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.
Claims
1. A diabetes risk prediction system based on multiple models, It is characterized in that It includes a learning model building module, an attribute information recognition and extraction module, a directed graph model building module, a node information recognition and extraction module, a feature vector extraction module and a network model building module; The learning model construction module is used to train a deep learning model based on the obtained diabetes-related information to construct a diabetes deep learning model; The attribute information identification and extraction module is used to preconfigure attribute information identification and extraction rules. The diabetes deep learning model uses the attribute information identification and extraction rules to identify and extract multiple diabetes patient historical information, and identifies and extracts multiple basic information attributes and multiple information attribute directional relationships under different diabetes types. The information attribute directional relationship is a directional relationship of the first information attribute→the second information attribute; The directed graph model construction module is used to construct a directed graph model of the diabetes type under the corresponding diabetes type based on multiple basic information attributes and multiple information attribute directional relationships under different diabetes types and graph theory principles, and the attribute nodes whose number of basic information attribute nodes and information attribute directional relationships in the directed graph model of the diabetes type reaches the target number are used as priority attribute nodes, when the priority attribute node is the second information attribute node, its corresponding first information attribute node is also a priority attribute node, and the rest are used as non-priority attribute nodes; The node information identification and extraction module is used to identify and extract multiple diabetes patient history information of the corresponding diabetes type based on the pre-configured node information identification and extraction rules corresponding to each diabetes type directed graph model. The diabetes deep learning model uses the node information identification and extraction rules to identify and extract the historical information of multiple diabetes patients of the corresponding diabetes type. The specific content of the priority attribute node under the diabetes type directed graph model is first extracted from the historical information of each diabetes patient and filled into the corresponding priority attribute node, and then the specific content of the non-priority attribute node is extracted and filled into the corresponding non-priority attribute node, so as to obtain the diabetes type directed graph of each diabetes patient under each diabetes type; The feature vector extraction module is used to extract a directed feature vector in which both the first information attribute node and the second information attribute node are filled from the diabetes type directed graph of each diabetes patient under each diabetes type; The network model building module is used to use all directional feature vectors extracted from each diabetic patient under each diabetes type as an input sample set and the corresponding diabetes type as an output sample to train the convolutional neural network model to obtain a diabetes type neural network model; The diabetes risk prediction system also includes a diabetes risk prediction module, which is used to receive information of a patient to be estimated, obtain each diabetes type directed graph model, call the node information identification and extraction module for each diabetes type directed graph model, and the diabetes deep learning model uses the node information identification and extraction rules corresponding to the diabetes type directed graph model to identify and extract the information of the patient to be estimated, first extract the specific content of the priority attribute node under the diabetes type directed graph model from the information of the patient to be estimated, fill it into the corresponding priority attribute node, and then extract the specific content of the non-priority attribute node and fill it into the corresponding non-priority attribute node, so as to obtain each diabetes type of the patient to be estimated. A diabetes type directed graph is obtained, and then a feature vector extraction module is called to extract a directed feature vector in which the first information attribute node and the second information attribute node are both filled from each diabetes type directed graph. The directed feature vector extracted from the diabetes type directed graph is input into the diabetes type neural network model as a separate input sample set to output whether the patient to be estimated has the risk of the diabetes type corresponding to the diabetes type directed graph. If the answer is yes, risk information of the patient to be estimated having the diabetes type is issued; if the answer is no, risk information of the patient to be estimated not having the diabetes type is issued, thereby estimating whether the patient to be estimated has the risk of diabetes and the risk of which type of diabetes he has.
2. The diabetes risk prediction system based on multiple models as claimed in claim 1, It is characterized in that The diabetes risk estimation module is used to first extract the specific content of the priority attribute node under the diabetes type directed graph model from the information of the patient to be estimated and fill it into the corresponding priority attribute node, then extract the specific content of the non-priority attribute node and fill it into the corresponding non-priority attribute node and mark it as an uncertain node when the filled specific content is uncertain, so as to obtain the diabetes type directed graph under each diabetes type of the patient to be estimated, and then call the feature vector extraction module to extract from each diabetes type directed graph the first information attribute node and the second information attribute node that are both filled and do not contain marked uncertain nodes as a determined directed feature vector, and the uncertain nodes that are both filled and contain marked uncertain nodes as a to-be-determined directed feature vector. When the number of determined directed feature vectors extracted from at least one diabetes type directed graph reaches a set number, the determined directed feature vectors extracted from the diabetes type directed graph will be preferentially input as a separate input sample set into the diabetes type neural network model to output whether the patient to be estimated has the risk of the diabetes type corresponding to the diabetes type directed graph. If it is yes, the information that the patient to be estimated has the risk of the diabetes type and the diabetes risk estimation is completed will be issued; The diabetes risk estimation module is also used to issue a risk that the patient to be estimated does not have the type of diabetes when the answer is no, and display the to-be-determined directional feature vectors corresponding to the diabetes type directed graph that have not reached the set number and their original sources for manual correction so as to correct the to-be-determined directional feature vectors to determined directional feature vectors. When it is determined that the number of determined directional feature vectors corresponding to a corrected diabetes type directed graph reaches the set number, the determined directional feature vectors extracted from the corrected diabetes type directed graph are input as a separate input sample set into the diabetes type neural network model to output whether the patient to be estimated has the risk of the diabetes type corresponding to the corrected diabetes type directed graph. When the answer is yes, the risk of the patient to be estimated having the diabetes type and the information that the diabetes risk estimation is completed are issued. When the answer is no, the risk of the patient to be estimated not having the diabetes type are issued, and it is determined whether there are still to-be-determined directional feature vectors corresponding to the diabetes type directed graph that have not reached the set number and need to be corrected. When the answer is yes, wait for manual correction to continue. When the answer is no, the risk of the patient to be estimated not having any diabetes type and the information that the diabetes risk estimation is completed are issued.
3. The diabetes risk prediction system based on multiple models as claimed in claim 2, It is characterized in that The diabetes risk prediction module is used to call the node information identification and extraction module for each diabetes type directed graph model, first extract the specific content of the priority attribute node under the diabetes type directed graph model from the information of the patient to be estimated, and directly fill it into the corresponding priority attribute node; then extract the specific content of the non-priority attribute node, fill it into the corresponding non-priority attribute node, and mark it as an uncertain node when the specific content filled is uncertain. Specifically, first extract the specific content of the non-priority attribute node that has a directed relationship with the priority attribute node, fill it into the corresponding non-priority attribute node, and mark it as an uncertain node when the specific content filled is uncertain; then extract the specific content of the non-priority attribute node that has a directed relationship with the non-priority attribute node, fill it into the corresponding non-priority attribute node, and mark it as an uncertain node when the specific content filled is uncertain.
4. The diabetes risk prediction system based on multiple models as claimed in claim 3, It is characterized in that When the specific content to be filled is uncertain, it is marked as an uncertain node: the specific content of a non-priority attribute node has been filled, and new specific content of the non-priority attribute node is currently extracted. There is a situation where the new specific content and the filled specific content cannot be identified as consistent. In this case, this non-priority attribute node is an uncertain node and the source of the new specific content and the source of the filled specific content are recorded for manual correction.
5. The diabetes risk prediction system based on multiple models as claimed in claim 1, It is characterized in that The node information identification and extraction module is used to first extract the specific content of the priority attribute node under the directed graph model of the diabetes type from the historical information of each diabetic patient and directly fill it into the corresponding priority attribute node; then extract the specific content of the non-priority attribute node and fill it into the corresponding non-priority attribute node. Specifically, the specific content of the non-priority attribute node that forms a directed relationship with the priority attribute node is first extracted and filled into the corresponding non-priority attribute node, and then the specific content of the non-priority attribute node that forms a directed relationship with the non-priority attribute node is extracted and filled into the corresponding non-priority attribute node.
6. The diabetes risk prediction system based on multiple models as claimed in claim 5, It is characterized in that The node information identification and extraction module is used for a non-priority attribute node. The specific content of the non-priority attribute node has been filled. At present, new specific content of the non-priority attribute node is extracted. There is a situation where the new specific content and the filled specific content cannot be identified as consistent. In this case, this non-priority attribute node is an uncertain node, and the specific content of the non-priority attribute node is manually corrected without marking the non-priority attribute node as an uncertain node.
7. The diabetes risk prediction system based on multiple models as claimed in claim 1, It is characterized in that The diabetes-related information includes data in the medical field of diabetes, diet data, exercise data and sleep data corresponding to diabetes, medical data on diabetes in various hospitals and research data related to diabetic foot, symptom data, treatment drug data, diet data, exercise data and sleep data of diabetic patients, and diabetes information crawled from the Internet by crawler tools; The historical information of diabetic patients includes the type of diabetes suffered by diabetic patients, basic information of diabetic patients, symptom information of diabetic patients, examination information of diabetic patients and life information of diabetic patients. The basic information of diabetic patients includes the gender, age and weight of diabetic patients. The life information of diabetic patients includes the diet information, exercise information and sleep information of diabetic patients.
8. The diabetes risk prediction system based on multiple models as claimed in claim 1, It is characterized in that The information of the patient to be estimated includes the basic information of the patient to be estimated, the symptom information of the patient to be estimated, the examination information of the patient to be estimated and the living information of the patient to be estimated. The basic information of the patient to be estimated includes the gender, age and weight of the patient to be estimated, and the living information of the patient to be estimated includes the diet information, exercise information and sleep information of the patient to be estimated.
