Method and System for Implementing Data Analysis and Processing of Chronic Disease Quality Control Based on Big Data

Abstract

The present invention discloses a method and system for realizing data analysis and processing of chronic disease quality control based on big data, which relates to the field of artificial intelligence and includes: First, obtain a chronic disease quality control assessment model jointly trained by known and simulated chronic disease quality control instances. Then, obtain a brand-new chronic disease quality control instance, load it into the model to extract a characterization vector. By determining the target key features corresponding to the simulated chronic disease quality control state in the feature domain, and based on the deviation between the characterization vector and the key features, obtain the corresponding target simulated chronic disease quality control state, and adjust it to a brand-new chronic disease quality control state, so as to realize the inference of the brand-new state by using the chronic disease quality control assessment model, and provide a more accurate and effective analysis and processing means for chronic disease management.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular, to a method and system for realizing data analysis and processing of chronic disease quality control based on big data. Background Art

[0002] With the development of big data technology, the demand for chronic disease management is increasing day by day. Traditional methods for chronic disease quality control data analysis are mostly based on limited known case data and are difficult to effectively cope with the continuously emerging complex and diverse chronic disease situations. In the face of a new chronic disease quality control state, existing models lack the ability of accurate inference. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for realizing data analysis and processing of chronic disease quality control based on big data.

[0004] In a first aspect, an embodiment of the present invention provides a method for realizing data analysis and processing of chronic disease quality control based on big data, including:

[0005] Obtain a chronic disease quality control evaluation model jointly trained by known chronic disease quality control instances and simulated chronic disease quality control instances, where the chronic disease quality control evaluation model is used to infer simulated chronic disease quality control states and known chronic disease quality control states;

[0006] Obtain a new chronic disease quality control instance, where the new chronic disease quality control state corresponding to the new chronic disease quality control instance is different from any of the known chronic disease quality control states;

[0007] Load the new chronic disease quality control instance into the chronic disease quality control evaluation model, and use the chronic disease quality control evaluation model to extract a feature vector of the new chronic disease quality control instance;

[0008] Determine target key features corresponding to each simulated chronic disease quality control state in the feature domain, and obtain a target simulated chronic disease quality control state corresponding to the new chronic disease quality control instance according to the deviation between the feature vector and each target key feature;

[0009] Adjust the target simulated chronic disease quality control state to the new chronic disease quality control state, so as to infer the new chronic disease quality control state through the chronic disease quality control evaluation model.

[0010] In a second aspect, an embodiment of the present invention provides a readable storage medium, where the readable storage medium includes a computer program, and when the computer program runs, it controls a computer device where the readable storage medium is located to execute the method described in the first aspect.

[0011] Compared with the prior art, the beneficial effects provided by the present invention include: By adopting a method and system for analyzing and processing chronic disease quality control data based on big data disclosed by the present invention, a chronic disease quality control evaluation model jointly trained by known and simulated chronic disease quality control instances is obtained. Then, a brand-new chronic disease quality control instance is obtained and loaded into the model to extract a characterization vector. By determining the target key features corresponding to the simulated chronic disease quality control state in the feature domain, based on the deviation between the characterization vector and the key features, the corresponding target simulated chronic disease quality control state is obtained and adjusted to a brand-new chronic disease quality control state, realizing the inference of the brand-new state using the chronic disease quality control evaluation model, and providing a more accurate and effective analysis and processing means for chronic disease management. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0013] Figure 1 It is a schematic flowchart of the steps of the method for analyzing and processing chronic disease quality control data based on big data provided by the embodiments of the present invention;

[0014] Figure 2 It is a schematic block diagram of the structure of the device for analyzing and processing chronic disease quality control data based on big data provided by the embodiments of the present invention;

[0015] Figure 3 It is a schematic block diagram of the structure of the computer device provided by the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated in the drawings here can be arranged and designed in various different configurations.

[0017] The following will describe in detail the specific embodiments of the present invention in conjunction with the accompanying drawings.

[0018] To solve the technical problems in the foregoing background art, Figure 1 It is a schematic flowchart of the method for analyzing and processing chronic disease quality control data based on big data provided by the embodiments of the present disclosure. The following will introduce in detail the method for analyzing and processing chronic disease quality control data based on big data.

[0019] Step S201: Obtain a chronic disease quality control assessment model jointly trained by known chronic disease quality control instances and simulated chronic disease quality control instances, where the chronic disease quality control assessment model is used to infer the simulated chronic disease quality control status and the known chronic disease quality control status;

[0020] Step S202: Obtain a brand-new chronic disease quality control instance, where the brand-new chronic disease quality control status corresponding to the brand-new chronic disease quality control instance is different from any of the known chronic disease quality control statuses;

[0021] Step S203: Load the brand-new chronic disease quality control instance into the chronic disease quality control assessment model, and use the chronic disease quality control assessment model to extract the feature vector of the brand-new chronic disease quality control instance;

[0022] Step S204: Determine the target key features corresponding to each simulated chronic disease quality control status in the feature domain, and obtain the target simulated chronic disease quality control status corresponding to the brand-new chronic disease quality control instance according to the deviation between the feature vector and each target key feature;

[0023] Step S205: Adjust the target simulated chronic disease quality control status to the brand-new chronic disease quality control status, so as to infer the brand-new chronic disease quality control status through the chronic disease quality control assessment model.

[0024] In an embodiment of the present invention, exemplarily, in a large medical system, the server is responsible for collecting known chronic disease quality control instance data from various medical institutions. These data cover a large number of case information of patients with diagnosed chronic diseases, including patient basic information (such as age, gender, family medical history, etc.), disease diagnosis data (disease type, duration of illness, etc.), treatment process records (medication conditions, inspection reports, etc.), and the corresponding known chronic disease quality control status, such as stable condition, fluctuating condition, etc. At the same time, the server uses a specific algorithm to construct simulated chronic disease quality control instances based on these known chronic disease quality control instances. For example, by randomly combining or slightly perturbing the features of known instances, simulated data with similar but not exactly the same features are generated, and the simulated chronic disease quality control status corresponding to these simulated instances is also set. The server uses these known chronic disease quality control instances and simulated chronic disease quality control instances to jointly train the initial recognition model. During the training process, the initial recognition model will learn to extract features from the instance data and try to infer the corresponding chronic disease quality control status. For example, for a known chronic disease quality control instance, the model will infer the first confidence level of this instance compared to all chronic disease quality control statuses based on its feature vector, and calculate the first basic error by comparing with the known target known chronic disease quality control status. At the same time, according to the first confidence level, determine the first inference confidence level compared to the remaining chronic disease quality control statuses except the target known chronic disease quality control status, and then determine the first simulation error compared to the target simulated chronic disease quality control status. For simulated chronic disease quality control instances, a similar inference process is also used to calculate the second error. By comprehensively combining the first error and the second error to construct an objective error function, the initial recognition model is optimized through multiple rounds of cycling. After each optimization, adjust the key features corresponding to each chronic disease quality control status in the feature domain according to the error backpropagation mechanism of the optimization direction. After multiple rounds of training, when the preset training termination state is reached, the final chronic disease quality control assessment model is obtained, and this model has the ability to accurately infer the simulated chronic disease quality control status and the known chronic disease quality control status. During the daily medical data update process, the server receives a new chronic disease case data uploaded from a primary hospital. The patient of this case has some unique symptom and medical history features, and its corresponding chronic disease quality control status is different from all previous known chronic disease quality control statuses, which constitutes a new chronic disease quality control instance. For example, the patient has a rare chronic disease complication that has never appeared in previous case data with a similar quality control status. The server loads the obtained new chronic disease quality control instance data into the trained chronic disease quality control assessment model. The model starts to analyze these data and extracts a feature vector that can represent the characteristics of this instance from the patient's various information (such as detailed symptom descriptions, special test indicators, etc.). This is like extracting the most critical information points from a complex report, and the vector composed of these information points will be used for subsequent comparison and analysis with the key features.For example, the model may extract elements such as the severity index of patient-specific symptoms and the abnormality degree of specific test indexes as elements of the characterization vector. The server stores the target key features corresponding to each simulated chronic disease quality control state in the feature domain. These target key features were determined during the previous training process of the chronic disease quality control assessment model and represent the typical feature combinations of each simulated chronic disease quality control state. For example, for a simulated "potential disease deterioration" quality control state, its target key features may include the aggravation trend of specific symptoms, the change of critical values of certain test indexes, etc. The server calculates the deviation between the characterization vector of the new chronic disease quality control instance and each target key feature. Here, the deviation can be measured by calculating similarity, etc., and the higher the similarity, the smaller the deviation. For example, through a specific similarity calculation algorithm, the similarity between the characterization vector and each target key feature is obtained respectively. The simulated chronic disease quality control state mapped by the target key feature corresponding to the highest similarity value is determined as the target simulated chronic disease quality control state corresponding to the new chronic disease quality control instance. Suppose that after calculation, the target key feature corresponding to the "potential disease deterioration" simulated chronic disease quality control state has the highest similarity with the characterization vector of the new chronic disease quality control instance, then "potential disease deterioration" is determined as the target simulated chronic disease quality control state. The server adjusts the determined target simulated chronic disease quality control state to the new chronic disease quality control state corresponding to the new chronic disease quality control instance. This adjustment process means that in the knowledge base of the chronic disease quality control assessment model, an association and identification path is established for this new quality control state. For example, the previously determined "potential disease deterioration" simulated chronic disease quality control state is marked as a specific new chronic disease quality control state corresponding to the new chronic disease quality control instance, such as the "special change of the disease condition under rare complications" state. After such adjustment, when a chronic disease quality control instance with similar features is input into the model next time, the model can identify and infer this new chronic disease quality control state, further enriching the model's recognition ability and support for chronic disease management. Through the above detailed steps and scenario examples, the method for realizing chronic disease quality control data analysis and processing based on big data is specifically implemented on the server side, providing a solid technical foundation for more accurate and comprehensive management of chronic diseases.

[0025] In an embodiment of the present invention, the following method is also provided.

[0026] Obtain the chronic disease quality control data to be processed;

[0027] Load the chronic disease quality control data to be processed into the chronic disease quality control assessment model, and use the chronic disease quality control assessment model to extract the characterization vector of the chronic disease quality control data to be processed;

[0028] Based on the feature vector of the chronic disease quality control data to be processed, infer the chronic disease quality control data to be processed to determine the target chronic disease quality control status corresponding to the chronic disease quality control data to be processed, where the target chronic disease quality control status is any of the known chronic disease quality control statuses or the new chronic disease quality control status.

[0029] In an embodiment of the present invention, for example, in a medical big data management system, the server undertakes the important task of processing a large amount of chronic disease-related data. The server obtains the chronic disease quality control data to be processed. For example, a number of community hospitals and specialized hospitals in a certain area regularly upload the chronic disease-related information of patients to the server, and this information constitutes the chronic disease quality control data to be processed. Among them, there is the information of a diabetic patient, including recent blood glucose monitoring data, glycated hemoglobin index, types and dosages of medications, records of diet and exercise habits, etc. Then, the server loads this chronic disease quality control data to be processed into the chronic disease quality control assessment model. The model is like an intelligent analysis engine and starts to deeply analyze the data to extract the feature vector. Taking the data of this diabetic patient as an example, the model may extract key elements such as the frequency of blood glucose fluctuations, the degree of deviation of glycated hemoglobin from the normal range, and the degree of dependence on specific medications as the feature vector. These elements comprehensively reflect the characteristic information of the chronic disease of this patient. After that, the server infers the data based on the extracted feature vector of the chronic disease quality control data to be processed. The model will compare this feature vector with the features corresponding to the known chronic disease quality control status and the new chronic disease quality control status. Suppose there are statuses such as "good blood glucose control" and "large blood glucose fluctuations, need to adjust the treatment plan" in the known chronic disease quality control status. If the model finds through comparison that the frequency of blood glucose fluctuations of this patient is high and the glycated hemoglobin deviates significantly from the normal range, and the feature vector has a high degree of matching with the features of the "large blood glucose fluctuations, need to adjust the treatment plan" status, it will determine that the target chronic disease quality control status corresponding to the chronic disease quality control data to be processed is "large blood glucose fluctuations, need to adjust the treatment plan". If the situation of this patient is special and its feature vector is more in line with the features of the new chronic disease quality control status (such as a new status determined due to rare complications before), the server will determine the target chronic disease quality control status as this new chronic disease quality control status, providing a basis for subsequent precise treatment and management.

[0030] In an embodiment of the present invention, the chronic disease quality control assessment model is obtained according to the following process.

[0031] Obtain known chronic disease quality control instances and simulated chronic disease quality control instances, where the simulated chronic disease quality control instances are constructed based on the known chronic disease quality control instances;

[0032] Infer the known chronic disease quality control instance using the initial recognition model, and determine the first error of the known chronic disease quality control instance compared to the first target chronic disease quality control state according to the first inference value generated by the inference. The first target chronic disease quality control state includes the target known chronic disease quality control state corresponding to the known chronic disease quality control instance and the target simulated chronic disease quality control state corresponding to the known chronic disease quality control instance;

[0033] Infer the simulated chronic disease quality control instance using the initial recognition model, and determine the second error of the simulated chronic disease quality control instance compared to the second target chronic disease quality control state according to the second inference value generated by the inference. The second target chronic disease quality control state includes the target known chronic disease quality control state and the target simulated chronic disease quality control state corresponding to the simulated chronic disease quality control instance respectively;

[0034] Construct a target error function according to the first error and the second error, and use the target error function to cyclically optimize the initial recognition model until the preset training termination state is met, and obtain the final chronic disease quality control evaluation model.

[0035] In an embodiment of the present invention, exemplarily, the server first obtains known chronic disease quality control instances and simulated chronic disease quality control instances. For example, the server collects a large number of hypertension patient cases from multiple hospitals as known chronic disease quality control instances, which include information such as patient age, blood pressure value, medication situation, complications, etc. and the corresponding known chronic disease quality control status, such as stable blood pressure, unstable blood pressure, etc. At the same time, the server constructs simulated chronic disease quality control instances based on these known instances. For example, it slightly adjusts the blood pressure fluctuation range of some patients or changes the medication dose combination to generate simulated instances. Then, the server uses the initial recognition model to infer the known chronic disease quality control instances. Taking a specific hypertension patient as an example, the initial recognition model extracts a feature vector based on the patient's various data and infers the first confidence level of this patient compared to all chronic disease quality control statuses. Assuming that the patient is actually in the target known chronic disease quality control status of "stable blood pressure", the model determines the first basic error of this patient compared to the "stable blood pressure" status according to the first confidence level. At the same time, through a series of operations, the first inference confidence level of this patient compared to other statuses except "stable blood pressure" is determined, and then the first simulation error compared to the target simulated chronic disease quality control status is obtained. By combining the first basic error and the first simulation error, the first error of this known chronic disease quality control instance compared to the first target chronic disease quality control status is determined. For the simulated chronic disease quality control instance, the server also uses the initial recognition model to infer. For example, for a simulated hypertension patient instance, the model first extracts its feature vector and obtains the second confidence level compared to all chronic disease quality control statuses. The known instance based on which this simulated instance is constructed is determined, and then its target simulated chronic disease quality control status is clarified. According to the second confidence level, the second basic error compared to the target simulated chronic disease quality control status is calculated, and then the second simulation error compared to the target known chronic disease quality control status is determined through similar operations as above. Finally, the second error compared to the second target chronic disease quality control status is obtained. Finally, the server constructs a target error function based on the first error and the second error. This function is like a ruler for measuring the accuracy of the model. The server uses it to perform cyclic optimization on the initial recognition model. In each round of optimization, the model continuously adjusts its own parameters and gradually improves the accuracy of inferring the chronic disease quality control status. When the performance of the model meets the preset training termination status, such as the error is reduced to a certain extent or the specified number of training rounds is reached, the server obtains the final chronic disease quality control evaluation model for accurately inferring the chronic disease quality control status in the future.

[0036] In an embodiment of the present invention, the cyclic optimization of the initial recognition model using the target error function can be implemented through the following example.

[0037] The initial recognition model is optimized in multiple rounds of loops using the target error function. After each round of loop optimization, the optimization direction corresponding to this round of loop optimization is determined, and according to the error backpropagation mechanism of the optimization direction, the key features corresponding to each chronic disease quality control state in the feature domain are adjusted; among them, the key features corresponding to each chronic disease quality control state obtained at the end of the last round of loop optimization are the target key features of each chronic disease quality control state.

[0038] In an embodiment of the present invention, exemplarily, the server starts to optimize the initial recognition model using the target error function. Taking the chronic disease quality control data of diabetic patients as an example, in the first round of optimization, the server constructs a target error function based on the errors between the previously calculated known chronic disease quality control instances and the simulated chronic disease quality control instances. This function reflects the gap between the current inference result of the model and the actual chronic disease quality control state. After each round of loop optimization, the server determines the optimization direction corresponding to this round of loop optimization. For example, after the first round of optimization, the server analyzes and finds that the model has a large error when judging whether a patient is in the chronic disease quality control state of "poor blood sugar control but no need to adjust the treatment plan". After in-depth analysis, it is determined that it is necessary to strengthen the consideration of relevant features such as the amplitude of blood sugar fluctuations and the change trend of glycated hemoglobin, which clarifies the optimization direction. Then, the server adjusts the key features corresponding to each chronic disease quality control state in the feature domain according to the error backpropagation mechanism of this optimization direction. For example, for the state of "poor blood sugar control but no need to adjust the treatment plan", in the feature domain, the weight of the amplitude of blood sugar fluctuations in the key features is increased, and at the same time, the measurement method of the change trend of glycated hemoglobin is slightly adjusted. In this way, the key features are more in line with the actual situation, which helps the model make more accurate inferences. As the loop optimization progresses, each round will continuously adjust the key features according to the error and optimization direction of the previous round. For example, in the second round of optimization, it is found that the model has a deviation when judging the state of "good blood sugar control", and the server will adjust the key features corresponding to the state of "good blood sugar control" according to this situation, such as further clarifying the manifestation method of the boundary value of the normal blood sugar range in the features. After multiple rounds of loop optimization, at the end of the last round of loop optimization, the key features corresponding to each chronic disease quality control state obtained by the server become the target key features of each chronic disease quality control state. These target key features are a set of features that can most accurately reflect each chronic disease quality control state after the model has been repeatedly adjusted and optimized. For example, for the state of "stable diabetes condition and good blood sugar control", the target key features accurately cover various factors such as the stable range of blood sugar values, the reasonable range of glycated hemoglobin, and the maintenance dose of specific drugs, enabling the model to be more accurate in subsequent inferences of chronic disease quality control states.

[0039] In an embodiment of the present invention, the step of inferring the known chronic disease quality control instance using the initial recognition model and determining a first error of the known chronic disease quality control instance compared to the first target chronic disease quality control state based on the first inference value generated by the inference can be implemented through the following examples.

[0040] Use the initial recognition model to extract the feature vector of the known chronic disease quality control instance, and perform inference based on the feature vector of the known chronic disease quality control instance to obtain a first confidence level of the known chronic disease quality control instance compared to all chronic disease quality control states;

[0041] Based on the first confidence level and the target known chronic disease quality control state corresponding to the known chronic disease quality control instance, determine a first basic error of the known chronic disease quality control instance compared to the target known chronic disease quality control state;

[0042] Based on the first confidence level, determine a first inference confidence level of the known chronic disease quality control instance compared to the remaining chronic disease quality control states other than the target known chronic disease quality control state;

[0043] Based on the first inference confidence level, determine a first simulation error of the known chronic disease quality control instance compared to the corresponding target simulated chronic disease quality control state;

[0044] Based on the first basic error and the first simulation error, determine a first error of the known chronic disease quality control instance compared to the first target chronic disease quality control state.

[0045] In an embodiment of the present invention, by way of example, first, the server uses the initial recognition model to extract the feature vector of the known chronic disease quality control instance. For a specific patient with coronary heart disease, the initial recognition model extracts a set of feature vectors from their medical records, such as age, family history, blood pressure, blood lipids, electrocardiogram data, and past treatment methods, etc., which can represent the patient's disease characteristics. Then, based on this feature vector, an inference is made to obtain a first confidence level of the known chronic disease quality control instance compared to all chronic disease quality control states. For example, the confidence level that the patient is in a "stable condition" is 0.7, the confidence level that the patient is in a "potential risk of disease deterioration" state is 0.2, and the confidence level that the patient is in an "acute onset of disease" state is 0.1, etc. Then, based on the first confidence level and the target known chronic disease quality control state corresponding to the patient, a first basic error is determined. Suppose the actual target known chronic disease quality control state of this patient is "stable condition", and the confidence level inferred by the model as "stable condition" is 0.7. Through a specific error calculation method (for example, setting the ideal state confidence level as 1 and taking the square of the difference between the actual confidence level and 1), the first basic error of this patient compared to the target known chronic disease quality control state of "stable condition" can be calculated, such as (1 - 0.7)2 = 0.09. Next, determine the first inference confidence level of this patient compared to the remaining chronic disease quality control states other than "stable condition". Based on the confidence levels of each state obtained previously, excluding "stable condition", the sum of the confidence levels of states such as "potential risk of disease deterioration" and "acute onset of disease" is 0.3, which is the first inference confidence level. Then, according to the first inference confidence level, determine the first simulation error of this patient compared to the corresponding target simulated chronic disease quality control state. The server analyzes the key features corresponding to each simulated chronic disease quality control state in the feature domain, calculates the similarity between the patient's feature vector and these key features, and determines the target simulated chronic disease quality control state. Assume that the target simulated chronic disease quality control state is determined to be "simulated slight fluctuation of the disease", and the ideal confidence level corresponding to the state of "simulated slight fluctuation of the disease" is set to 0.8, and the current first inference confidence level is 0.3. According to a similar error calculation method, the first simulation error is obtained, such as (0.8 - 0.3) 2 = 0.25. Finally, based on the first base error of 0.09 and the first simulation error of 0.25, determine the first error of this known chronic disease quality control instance compared to the first target chronic disease quality control state (including the target known chronic disease quality control state "stable condition" and the target simulated chronic disease quality control state "simulated slight fluctuation of the disease"), for example, add the two together, 0.09 + 0.25 = 0.34. This first error reflects the comprehensive deviation degree of the model's inference for this known chronic disease quality control instance and is used for subsequent optimization of the initial recognition model.

[0046] In an embodiment of the present invention, the determining the first inference confidence level of the known chronic disease quality control instance compared to the remaining chronic disease quality control states other than the target known chronic disease quality control state according to the first confidence level can be implemented through the following example.

[0047] According to the target known chronic disease quality control state corresponding to the known chronic disease quality control instance, obtain the state identification vector corresponding to the known chronic disease quality control instance, and the number of elements of the state identification vector is the same as the number of preset chronic disease quality control states;

[0048] Perform a logical complement generation operation on the state identification vector corresponding to the known chronic disease quality control instance to obtain the complementary state vector corresponding to the known chronic disease quality control instance;

[0049] According to the first confidence level and the complementary state vector corresponding to the known chronic disease quality control instance, determine the first inference confidence level of the known chronic disease quality control instance compared to the remaining chronic disease quality control states other than the target known chronic disease quality control state.

[0050] In an embodiment of the present invention, by way of example, take the server processing the chronic disease quality control data of hypertensive patients as an example. Assume that there are three preset chronic disease quality control states: "stable blood pressure", "fluctuating but controllable blood pressure", and "uncontrolled blood pressure". For a specific hypertensive patient, this is a known chronic disease quality control instance, and its target known chronic disease quality control state is "stable blood pressure". The server first obtains the corresponding state identification vector according to this target known chronic disease quality control state. Since there are three preset states, the state identification vector has three elements, corresponding to "stable blood pressure", "fluctuating but controllable blood pressure", and "uncontrolled blood pressure". Because the target state of this patient is "stable blood pressure", the state identification vector is [1, 0, 0], where 1 indicates that the patient is in this state and 0 indicates not. Next, the server performs a logical complement generation operation on this state identification vector [1, 0, 0]. The logical complement is to change the element 0 in the vector to 1 and 1 to 0, thus obtaining the complementary state vector [0, 1, 1]. This complementary state vector represents other chronic disease quality control states except "stable blood pressure" (the target known chronic disease quality control state). Then, the server combines the first confidence level to determine the first inference confidence level. Previously, using the initial recognition model, a feature vector was extracted based on the patient's medical record data, such as blood pressure measurement values, medication conditions, living habits, etc., and the first confidence level of this patient compared to all chronic disease quality control states was inferred, assumed to be [0.8, 0.1, 0.1], corresponding to the confidence levels of "stable blood pressure", "fluctuating but controllable blood pressure", and "uncontrolled blood pressure" respectively. The server performs an element-wise multiplication operation on the first confidence level vector [0.8, 0.1, 0.1] and the complementary state vector [0, 1, 1], that is, 0.8×0 + 0.1×1 + 0.1×1 = 0.2. This result 0.2 is the first inference confidence level of this known chronic disease quality control instance compared to the remaining chronic disease quality control states ("fluctuating but controllable blood pressure" and "uncontrolled blood pressure") except "stable blood pressure" (the target known chronic disease quality control state). It reflects the overall confidence level of the model that the patient is in other states except the target known chronic disease quality control state, and is used to calculate the error of the model's inference for this patient in the subsequent process, thereby optimizing the initial recognition model.

[0051] In an embodiment of the present invention, to determine the first simulation error of the known chronic disease quality control instance compared to the corresponding target simulated chronic disease quality control state according to the first inference confidence level, the following example can be executed.

[0052] Obtain the target simulated chronic disease quality control state corresponding to the known chronic disease quality control instance;

[0053] Determine the first simulation error of the known chronic disease quality control instance compared to the target simulated chronic disease quality control state according to the first inference confidence level and the target value of the simulated chronic disease quality control state corresponding to the target simulated chronic disease quality control state.

[0054] In an embodiment of the present invention, by way of example, still taking the chronic disease quality control data processing of hypertensive patients as an example, the server continues to perform related operations. First, the server needs to obtain the target simulated chronic disease quality control state corresponding to the known chronic disease quality control instance. Assume that the server stores a simulated chronic disease quality control instance library constructed based on a large number of known hypertensive chronic disease instances, and the key features corresponding to each simulated chronic disease quality control state. The server determines the target simulated chronic disease quality control state by calculating the similarity between the feature vector of the known hypertensive patient instance and the key features of each simulated chronic disease quality control state. For example, the server calculates and finds that the similarity between the feature vector of this patient and the key features corresponding to the "simulated short-term blood pressure fluctuation state" is the highest, so the "simulated short-term blood pressure fluctuation state" is determined as the target simulated chronic disease quality control state. Next, according to the first inference confidence and the target value of the simulated chronic disease quality control state corresponding to the target simulated chronic disease quality control state, the first simulation error is determined. Assume that the first inference confidence calculated previously is 0.3, and this value reflects the possibility that the model believes that the patient is in a state other than the target known chronic disease quality control state (such as "stable blood pressure"). For the "simulated short-term blood pressure fluctuation state", the server has preset a target value of the simulated chronic disease quality control state to measure the ideal degree of the model's judgment of this state. Assume that this target value is 0.7, indicating that ideally the confidence of the model in judging that the patient is in this simulated state should be 0.7. The server determines the first simulation error through a specific error calculation method. For example, use the square of the difference between the two to calculate, that is, (0.7 - 0.3) 2 = 0.16. This 0.16 is the first simulation error of the known chronic disease quality control instance compared to the target simulated chronic disease quality control state of the "simulated short-term blood pressure fluctuation state". This error reflects the deviation of the matching degree between the model and the target simulated chronic disease quality control state for this patient. Together with the first basic error, etc., it provides a basis for the server to optimize the initial recognition model, making the model more accurate in subsequent inferences of chronic disease quality control states.

[0055] In an embodiment of the present invention, the obtaining of the target simulated chronic disease quality control state corresponding to the known chronic disease quality control instance can be implemented through the following example.

[0056] Determine the first key features corresponding to each simulated chronic disease quality control state in the feature domain, where the first key features are the key features corresponding to each simulated chronic disease quality control state in this round of training;

[0057] Respectively determine the similarity between the feature vector of the known chronic disease quality control instance and each first key feature;

[0058] Map the simulated chronic disease quality control state corresponding to the highest similarity value among multiple said similarity values to the target simulated chronic disease quality control state corresponding to the known chronic disease quality control instance.

[0059] In an embodiment of the present invention, exemplarily, in this round of training, the server first determines the first key features corresponding to each simulated chronic disease quality control state in the feature domain. For example, there are three simulated chronic disease quality control states preset in the server: "simulated mild blood glucose fluctuation", "simulated moderate blood glucose fluctuation", and "simulated severe blood glucose fluctuation". For the state of "simulated mild blood glucose fluctuation", its first key features may include the fluctuation frequency of blood glucose values within a certain range, minor changes in glycated hemoglobin, etc.; the first key features of "simulated moderate blood glucose fluctuation" involve an increase in the amplitude of blood glucose fluctuations, slight changes in the effects of specific hypoglycemic drugs, etc.; the first key features of "simulated severe blood glucose fluctuation" include specific index combinations such as sharp changes in blood glucose and the possibility of multiple complications. Then, the server respectively determines the similarity between the feature vector of the known chronic disease quality control instance (a certain diabetic patient) and each first key feature. The server extracts the feature vector from information such as the patient's recent blood glucose monitoring data, medication records, and changes in physical indicators. Then, through a specific similarity calculation algorithm, the feature vector is compared with the first key feature of "simulated mild blood glucose fluctuation" to obtain a similarity value, assumed to be 0.6; compared with the first key feature of "simulated moderate blood glucose fluctuation", the similarity value is 0.8; compared with the first key feature of "simulated severe blood glucose fluctuation", the similarity value is 0.4. Finally, the server maps the simulated chronic disease quality control state corresponding to the highest similarity value among multiple similarity values to the target simulated chronic disease quality control state corresponding to the known chronic disease quality control instance. In the above example, 0.8 is the highest value among the three similarity values, and it corresponds to the first key feature corresponding to the state of "simulated moderate blood glucose fluctuation". Therefore, the server determines "simulated moderate blood glucose fluctuation" as the target simulated chronic disease quality control state corresponding to the known chronic disease quality control instance. This determination process helps to calculate the simulation error subsequently, thereby optimizing the initial recognition model and making the model more accurate in judging the chronic disease quality control state of diabetic patients.

[0060] In an embodiment of the present invention, the use of the initial recognition model to infer the simulated chronic disease quality control instance and determine the second error of the simulated chronic disease quality control instance compared to the second target chronic disease quality control state based on the second inference value generated by the inference can be implemented through the following example.

[0061] Extract the feature vector of the simulated chronic disease quality control instance using the initial recognition model, and make an inference based on the feature vector of the simulated chronic disease quality control instance to obtain the second confidence level of the simulated chronic disease quality control instance compared to all chronic disease quality control states;

[0062] Determine at least one known chronic disease quality control instance used to construct the simulated chronic disease quality control instance, and determine the target simulated chronic disease quality control state corresponding to the simulated chronic disease quality control instance according to the target simulated chronic disease quality control state corresponding to the at least one known chronic disease quality control instance;

[0063] Determine the second basic error of the simulated chronic disease quality control instance compared to the target simulated chronic disease quality control state according to the second confidence level and the target simulated chronic disease quality control state corresponding to the simulated chronic disease quality control instance;

[0064] Determine the second inference confidence level of the simulated chronic disease quality control instance compared to the remaining chronic disease quality control states except the target simulated chronic disease quality control state according to the second confidence level;

[0065] Determine the second simulation error of the simulated chronic disease quality control instance compared to the corresponding target known chronic disease quality control state according to the second inference confidence level;

[0066] Determine the second error of the simulated chronic disease quality control instance compared to the second target chronic disease quality control state according to the second basic error and the second simulation error.

[0067] In an embodiment of the present invention, exemplarily, the server extracts the feature vector of the simulated chronic disease quality control instance using the initial recognition model. Assuming the situation of a heart disease patient is simulated, the server extracts the feature vector representing the characteristics of this simulated instance from the simulated patient medical record data, such as the simulated heart rate fluctuation range, the simulated cardiac function indicators, the simulated frequency of chest pain symptoms, etc. Then, based on this feature vector, an inference is made to obtain the second confidence level of this simulated chronic disease quality control instance compared to all chronic disease quality control states. For example, it is inferred that the confidence level of this simulated patient being in the "stable condition" state is 0.6, the confidence level of being in the "potential risk of disease deterioration" state is 0.3, the confidence level of being in the "acute onset of disease" state is 0.1, etc. Then, the server determines at least one known chronic disease quality control instance used to construct this simulated chronic disease quality control instance. Assuming this simulated instance is constructed based on three known heart disease patient instances, the server determines the target simulated chronic disease quality control state corresponding to this simulated chronic disease quality control instance according to the target simulated chronic disease quality control states corresponding to these three known chronic disease quality control instances. For example, if the target simulated chronic disease quality control states corresponding to these three known patients all tend to be "simulated slow progression of the disease" after analysis, then the server determines "simulated slow progression of the disease" as the target simulated chronic disease quality control state corresponding to this simulated chronic disease quality control instance. Next, according to the second confidence level and the target simulated chronic disease quality control state corresponding to this simulated chronic disease quality control instance, the second basic error is determined. Assuming the ideal confidence level for the "simulated slow progression of the disease" state is set to 0.8, and the confidence level corresponding to this state in the current second confidence level is 0.6, through a specific error calculation method (such as the square of the difference between the two), the second basic error can be obtained as (0.8 - 0.6) 2 = 0.04. After that, according to the second confidence level, the second inference confidence level of this simulated chronic disease quality control instance compared to the remaining chronic disease quality control states except the target simulated chronic disease quality control state is determined. The sum of the confidence levels of other states except "simulated slow progression of the disease" is 0.4, which is the second inference confidence level. Then, according to the second inference confidence level, the second simulation error of this simulated chronic disease quality control instance compared to the corresponding target known chronic disease quality control state is determined. Assuming the target known chronic disease quality control state is determined to be "stable condition", and the ideal confidence level corresponding to the "stable condition" state is set to 0.7, and the current second inference confidence level is 0.4, using the same error calculation method, the second simulation error is obtained as (0.7 - 0.4) 2= 0.09. Finally, based on the second baseline error of 0.04 and the second simulation error of 0.09, determine the second error of this simulated chronic disease quality control instance compared to the second target chronic disease quality control state (including the target simulated chronic disease quality control state "simulated slow progression of the disease" and the target known chronic disease quality control state "stable condition"), that is, 0.04 + 0.09 = 0.13. This second error is used to evaluate the accuracy of the initial recognition model's inference on the simulated chronic disease quality control instance and provide a basis for optimizing the model.

[0068] In an embodiment of the present invention, the determining of the second inference confidence of the simulated chronic disease quality control instance compared to the remaining chronic disease quality control states other than the target simulated chronic disease quality control state according to the second confidence can be executed through the following example.

[0069] Obtain a state identification vector corresponding to the simulated chronic disease quality control instance according to the target known chronic disease quality control state corresponding to the simulated chronic disease quality control instance;

[0070] Perform a logical complement generation operation on the state identification vector corresponding to the simulated chronic disease quality control instance to obtain a complementary state vector corresponding to the simulated chronic disease quality control instance;

[0071] Determine the first inference confidence of the known chronic disease quality control instance compared to the remaining chronic disease quality control states other than the target known chronic disease quality control state according to the second confidence and the complementary state vector corresponding to the simulated chronic disease quality control instance.

[0072] In an embodiment of the present invention, by way of example, the server first obtains a corresponding status identification vector according to the target known chronic disease quality control status corresponding to the simulated chronic disease quality control instance. Suppose the server presets four chronic heart disease quality control statuses, namely "stable condition", "slight fluctuation of condition", "moderate deterioration of condition", and "severe deterioration of condition". The target known chronic disease quality control status corresponding to this simulated chronic disease quality control instance is "stable condition", then the status identification vector has four elements corresponding to these four statuses. Since it is "stable condition", this vector is [1, 0, 0, 0], where 1 represents the target known chronic disease quality control status corresponding to this simulated instance, and 0 represents other statuses. Next, the server performs a logical complement generation operation on this status identification vector [1, 0, 0, 0]. The so-called logical complement means changing the element 0 in the vector to 1 and 1 to 0, thus obtaining a complementary status vector [0, 1, 1, 1]. This complementary status vector represents other chronic disease quality control statuses except "stable condition" (the target known chronic disease quality control status). Then, the server combines the second confidence level to determine the second inference confidence level. Previously, using the initial recognition model, a characterization vector was extracted based on the simulated medical record data of the simulated heart disease patient, such as the simulated changes in various heart indicators and the simulated symptom manifestations, etc., and the second confidence level of this simulated chronic disease quality control instance compared to all chronic disease quality control statuses was inferred, supposed to be [0.7, 0.1, 0.1, 0.1], corresponding to the confidence levels of "stable condition", "slight fluctuation of condition", "moderate deterioration of condition", and "severe deterioration of condition" respectively. The server multiplies the second confidence level vector [0.7, 0.1, 0.1, 0.1] with the complementary status vector [0, 1, 1, 1] for corresponding element multiplication, that is, (0.7×0 + 0.1×1 + 0.1×1 + 0.1×1 = 0.3). This result 0.3 is the second inference confidence level of this simulated chronic disease quality control instance compared to the remaining chronic disease quality control statuses ("slight fluctuation of condition", "moderate deterioration of condition", "severe deterioration of condition") except "stable condition" (the target known chronic disease quality control status). It reflects the overall confidence level of the model in the simulated instance being in other statuses except the target known chronic disease quality control status, and will be used later to calculate the error of the model's inference for this simulated instance, so as to further optimize the initial recognition model and make its judgment of the chronic heart disease quality control status more accurate.

[0073] In an embodiment of the present invention, the second simulation error of the simulated chronic disease quality control instance compared to the corresponding target known chronic disease quality control status can be determined according to the second inference confidence level through the following example implementation.

[0074] Obtain the target known chronic disease quality control status corresponding to the simulated chronic disease quality control instance;

[0075] Determine a second simulation error of the simulated chronic disease quality control instance relative to the target known chronic disease quality control status according to the second inference confidence level and a known chronic disease quality control status target value corresponding to the target known chronic disease quality control status.

[0076] In an embodiment of the present invention, by way of example, it is assumed that the server is processing a simulated hypertension chronic disease quality control instance, which simulates a virtual patient with a specific blood pressure fluctuation pattern, medication situation, and related symptoms. First, the server obtains the target known chronic disease quality control status corresponding to the simulated chronic disease quality control instance. The server stores instance data of numerous known hypertension patients and their corresponding quality control statuses. By comparing and analyzing the features of the simulated instance with those of the known instances, the target known chronic disease quality control status corresponding to the simulated instance is determined. For example, through analysis, it is found that the features of this simulated patient most closely match those of numerous patients in the known chronic disease quality control status of "good and stable blood pressure control", so "good and stable blood pressure control" is determined as the target known chronic disease quality control status. Next, the server determines the second simulation error according to the second inference confidence level and a known chronic disease quality control status target value corresponding to the target known chronic disease quality control status. It is assumed that through the inference of the simulated instance by the initial recognition model before, a second confidence distribution is obtained, and then the second inference confidence level is calculated to be 0.2. This means that the model believes that the probability of the simulated instance being in a state other than the target simulated chronic disease quality control status is 0.2. For the target known chronic disease quality control status of "good and stable blood pressure control", the server has preset a known chronic disease quality control status target value to represent the confidence level of the model's judgment that the simulated instance is in this state under ideal circumstances. Assume that this target value is 0.8. The server uses a specific error calculation method to determine the second simulation error. For example, the square of the difference between the two is used for calculation, that is, ((0.8 - 0.2) 2 = 0.36). This 0.36 is the second simulation error of the simulated chronic disease quality control instance relative to the target known chronic disease quality control status of "good and stable blood pressure control". This error reflects the deviation of the matching degree between the model's simulated instance and the target known chronic disease quality control status. Together with other errors, it will provide a key basis for the server to optimize the initial recognition model, prompting the model to be more accurate and reliable in subsequent inferences of the hypertension chronic disease quality control status.

[0077] In an embodiment of the present invention, the obtaining of the target known chronic disease quality control status corresponding to the simulated chronic disease quality control instance may be implemented through the following example.

[0078] Determine second key features corresponding to each known chronic disease quality control status in the feature domain, where the second key features are the key features corresponding to the known chronic disease quality control status in this round of training;

[0079] Determine the similarity between the representation vector of the simulated chronic disease quality control instance and each second key feature respectively;

[0080] Determine the known chronic disease quality control state mapped by the second key feature corresponding to the highest similarity value in each similarity as the target known chronic disease quality control state corresponding to the simulated chronic disease quality control instance.

[0081] In an embodiment of the present invention, by way of example, assume that the server is processing a chronic disease quality control instance simulating a diabetic patient, and this instance contains information such as simulated blood glucose data, glycated hemoglobin values, types and frequencies of medications, etc. First, the server determines the second key features corresponding to each known chronic disease quality control state in the feature domain. In this round of training, the known chronic disease quality control states for diabetes are "stable blood glucose and well-controlled", "slightly fluctuating blood glucose but controllable", "moderately fluctuating blood glucose requiring treatment adjustment", and "severely uncontrolled blood glucose". For the "stable blood glucose and well-controlled" state, its second key features may include the blood glucose value being stable within a specific normal range, glycated hemoglobin being close to the normal level, using stable hypoglycemic medications at a conventional dose, etc.; the second key features of "slightly fluctuating blood glucose but controllable" may be that the blood glucose fluctuates within a certain small range, glycated hemoglobin deviates slightly from the normal, and the medication adjustment range is small, etc.; the second key features of "moderately fluctuating blood glucose requiring treatment adjustment" involve a larger amplitude of blood glucose fluctuation, a significant deviation of glycated hemoglobin, and the need to replace or adjust the medication dose, etc.; the second key features of "severely uncontrolled blood glucose" are specific indicator combinations such as a sharp increase or decrease in blood glucose, a severe abnormality in glycated hemoglobin, and the occurrence of multiple complications. Then, the server respectively determines the similarity between the representation vector of the simulated chronic disease quality control instance and each second key feature. The server extracts the representation vector from the various simulated data of the simulated diabetic patient. Through a specific similarity calculation algorithm, the representation vector is compared with the second key feature of "stable blood glucose and well-controlled", and a similarity value is obtained, assumed to be 0.5; compared with the second key feature of "slightly fluctuating blood glucose but controllable", the similarity value is 0.7; compared with the second key feature of "moderately fluctuating blood glucose requiring treatment adjustment", the similarity value is 0.6; compared with the second key feature of "severely uncontrolled blood glucose", the similarity value is 0.2. Finally, the server maps the known chronic disease quality control state corresponding to the second key feature with the highest similarity value in each similarity to determine the target known chronic disease quality control state corresponding to the simulated chronic disease quality control instance. In the above example, 0.7 is the highest value among the four similarity values, and it corresponds to the second key feature corresponding to the "slightly fluctuating blood glucose but controllable" state. Therefore, the server determines "slightly fluctuating blood glucose but controllable" as the target known chronic disease quality control state corresponding to this simulated chronic disease quality control instance. This determination process provides an important basis for subsequent calculation of the simulation error and optimization of the initial recognition model, making the model more accurate in judging the chronic disease quality control state of diabetes.

[0082] In an embodiment of the present invention, the step of obtaining the target simulated chronic disease quality control state corresponding to the new chronic disease quality control instance according to the deviation between the representation vector and each target key feature can be implemented through the following example.

[0083] Determine the similarity between the characterization vector and each target key feature;

[0084] Map the simulated chronic disease quality control status corresponding to the target key feature with the highest similarity value in each similarity, and determine it as the target simulated chronic disease quality control status corresponding to the new chronic disease quality control instance.

[0085] In an embodiment of the present invention, for example, the server receives a new chronic disease quality control instance, which is the case of a chronic kidney disease patient, and the comprehensive situation such as its symptoms and examination indicators is different from the known chronic kidney disease quality control status in the past. The server has obtained a chronic disease quality control assessment model, which includes the target key features corresponding to each simulated chronic disease quality control status in the feature domain. First, the server determines the similarity between the characterization vector of the new chronic disease quality control instance and each target key feature. Suppose the server has simulated chronic disease quality control statuses such as "simulated slow decline of renal function", "simulated abnormal fluctuation of proteinuria", and "simulated accompanying abnormal blood pressure" for chronic kidney disease and their corresponding target key features. For "simulated slow decline of renal function", its target key features may include a specific change range of glomerular filtration rate, a slow upward trend of serum creatinine, etc.; the target key features of "simulated abnormal fluctuation of proteinuria" involve the fluctuation range of urinary protein quantification and the change of specific protein components, etc.; the target key features of "simulated accompanying abnormal blood pressure" include the degree to which the blood pressure value exceeds the normal range and the correlation between blood pressure fluctuation and renal function indicators. The server extracts the characterization vector from the case data of the new chronic kidney disease patient, such as information in the renal function inspection report, blood pressure monitoring record, urine analysis result, etc. Then, through a special similarity calculation algorithm, the characterization vector is compared with the target key features of "simulated slow decline of renal function", and the similarity is obtained as 0.6; compared with the target key features of "simulated abnormal fluctuation of proteinuria", the similarity is 0.8; compared with the target key features of "simulated accompanying abnormal blood pressure", the similarity is 0.5. Finally, the server maps the simulated chronic disease quality control status corresponding to the target key feature with the highest similarity value in each similarity, and determines it as the target simulated chronic disease quality control status corresponding to the new chronic disease quality control instance. In the above example, 0.8 is the highest value among the three similarity values, and it corresponds to the target key features of the "simulated abnormal fluctuation of proteinuria" state. Therefore, the server determines "simulated abnormal fluctuation of proteinuria" as the target simulated chronic disease quality control status corresponding to the new chronic kidney disease quality control instance. In this way, the server can find the most matching simulated chronic disease quality control status based on the deviation between the characterization vector of the new chronic disease quality control instance and each target key feature, laying a foundation for the subsequent analysis and processing of this new situation.

[0086] Please refer to Figure 2 , Figure 2The device 110 for realizing the data analysis and processing of chronic disease quality control based on big data provided by the embodiments of the present invention includes:

[0087] An acquisition module 1101, configured to acquire a chronic disease quality control evaluation model obtained by jointly training known chronic disease quality control instances and simulated chronic disease quality control instances, where the chronic disease quality control evaluation model is used to infer simulated chronic disease quality control states and known chronic disease quality control states; acquire a brand-new chronic disease quality control instance, and the brand-new chronic disease quality control state corresponding to the brand-new chronic disease quality control instance is different from any of the known chronic disease quality control states;

[0088] An analysis module 1102, configured to load the brand-new chronic disease quality control instance into the chronic disease quality control evaluation model, and use the chronic disease quality control evaluation model to extract a feature vector of the brand-new chronic disease quality control instance; determine target key features corresponding to each simulated chronic disease quality control state in the feature domain, and obtain a target simulated chronic disease quality control state corresponding to the brand-new chronic disease quality control instance according to the deviation between the feature vector and each target key feature; adjust the target simulated chronic disease quality control state to the brand-new chronic disease quality control state, so as to infer the brand-new chronic disease quality control state through the chronic disease quality control evaluation model.

[0089] It should be noted that the implementation principle of the foregoing device 110 for realizing the data analysis and processing of chronic disease quality control based on big data can refer to the implementation principle of the foregoing method for realizing the data analysis and processing of chronic disease quality control based on big data, which will not be elaborated here. It should be understood that the division of each module of the above device is only a logical function division. In actual implementation, it can be fully or partially integrated into a physical entity, or physically separated. And these modules can all be implemented in the form of software called by a processing element; they can also all be implemented in the form of hardware; or some modules can be implemented in the form of software called by a processing element, and some modules can be implemented in the form of hardware. For example, the device 110 for realizing the data analysis and processing of chronic disease quality control based on big data can be a separately established processing element, or can be integrated in a certain chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a certain processing element of the above device to perform the functions of the device 110 for realizing the data analysis and processing of chronic disease quality control based on big data. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or independently implemented. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed by the integrated logic circuit in the processor element or the instruction in the form of software.

[0090] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as: one or more application specific integrated circuits (ASICs), or, one or more digital signal processors (DSPs), or, one or more field programmable gate arrays (FPGAs), etc. Again, when a certain module above is implemented in the form of a processing element scheduler code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processors that can call program code. Again, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0091] An embodiment of the present invention provides a computer device 100. The computer device 100 includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the foregoing apparatus 110 for implementing chronic disease quality control data analysis and processing based on big data. As Figure 3 shown, Figure 3 is a structural block diagram of the computer device 100 provided by an embodiment of the present invention. The computer device 100 includes an apparatus 110 for implementing chronic disease quality control data analysis and processing based on big data, a memory 111, a processor 112, and a communication unit 113. To achieve data transmission or interaction, the elements of the memory 111, the processor 112, and the communication unit 113 are directly or indirectly electrically connected to each other.

[0092] For illustrative purposes, the foregoing description has been made with reference to specific embodiments. However, the above illustrative discussion is not intended to be exhaustive or to limit the disclosure to the precise forms disclosed. Numerous modifications and variations are possible in light of the above teachings. The embodiments are chosen and described in order to best illustrate the principles of the disclosure and its practical applications, so that those skilled in the art can best utilize the disclosure and utilize various embodiments with different modifications to suit the particular applications contemplated.

Claims

1. A method for realizing data analysis and processing of chronic disease quality control based on big data, characterized in that, Including: Obtain a chronic disease quality control assessment model jointly trained by known chronic disease quality control instances and simulated chronic disease quality control instances, where the chronic disease quality control assessment model is used to infer simulated chronic disease quality control states and known chronic disease quality control states; Obtain a brand-new chronic disease quality control instance, where the brand-new chronic disease quality control state corresponding to the brand-new chronic disease quality control instance is different from any of the known chronic disease quality control states; Load the brand-new chronic disease quality control instance into the chronic disease quality control assessment model, and use the chronic disease quality control assessment model to extract the feature vector of the brand-new chronic disease quality control instance; Determine the target key features corresponding to each simulated chronic disease quality control state in the feature domain, and obtain the target simulated chronic disease quality control state corresponding to the brand-new chronic disease quality control instance according to the deviation between the feature vector and each target key feature; Adjust the target simulated chronic disease quality control state to the brand-new chronic disease quality control state, so as to infer the brand-new chronic disease quality control state through the chronic disease quality control assessment model; The chronic disease quality control assessment model is obtained according to the following process, including: Obtain known chronic disease quality control instances and simulated chronic disease quality control instances, where the simulated chronic disease quality control instances are constructed based on the known chronic disease quality control instances; Use the initial recognition model to extract the feature vector of the known chronic disease quality control instance, and make an inference based on the feature vector of the known chronic disease quality control instance to obtain the first confidence level of the known chronic disease quality control instance compared to all chronic disease quality control states; According to the first confidence level and the target known chronic disease quality control state corresponding to the known chronic disease quality control instance, determine the first basic error of the known chronic disease quality control instance compared to the target known chronic disease quality control state; According to the first confidence level, determine the first inference confidence level of the known chronic disease quality control instance compared to the remaining chronic disease quality control states other than the target known chronic disease quality control state; According to the first inference confidence level, determine the first simulation error of the known chronic disease quality control instance compared to the corresponding target simulated chronic disease quality control state; According to the first basic error and the first simulation error, determine the first error of the known chronic disease quality control instance compared to the first target chronic disease quality control state, where the first target chronic disease quality control state includes the target known chronic disease quality control state corresponding to the known chronic disease quality control instance and the target simulated chronic disease quality control state corresponding to the known chronic disease quality control instance; Use the initial recognition model to make an inference on the simulated chronic disease quality control instance, and determine the second error of the simulated chronic disease quality control instance compared to the second target chronic disease quality control state according to the second inference value generated by the inference, where the second target chronic disease quality control state includes the target known chronic disease quality control state and the target simulated chronic disease quality control state corresponding to the simulated chronic disease quality control instance respectively; Construct a target error function based on the first error and the second error, and use the target error function to perform multiple rounds of cyclic optimization on the initial recognition model. After each round of cyclic optimization, determine the optimization direction corresponding to this round of cyclic optimization, and adjust the key features corresponding to each chronic disease quality control state in the feature domain according to the error backpropagation mechanism of the optimization direction. Among them, the key features corresponding to each chronic disease quality control state obtained at the end of the last round of cyclic optimization are the target key features of each chronic disease quality control state, until the preset training termination state is met, and the final chronic disease quality control evaluation model is obtained.

2. The method according to claim 1, wherein The method further includes: Obtain the chronic disease quality control data to be processed; Load the chronic disease quality control data to be processed into the chronic disease quality control evaluation model, and use the chronic disease quality control evaluation model to extract the feature vector of the chronic disease quality control data to be processed; Infer the chronic disease quality control data to be processed according to the feature vector of the chronic disease quality control data to be processed, so as to determine the target chronic disease quality control state corresponding to the chronic disease quality control data to be processed, and the target chronic disease quality control state is any one of the known chronic disease quality control states or the new chronic disease quality control state.

3. The method according to claim 1, wherein The determining the first inference confidence of the known chronic disease quality control instance compared with the remaining chronic disease quality control states except the target known chronic disease quality control state according to the first confidence includes: According to the target known chronic disease quality control state corresponding to the known chronic disease quality control instance, obtain the state identification vector corresponding to the known chronic disease quality control instance, and the number of elements of the state identification vector is the same as the number of preset chronic disease quality control states; Perform a logical complement generation operation on the state identification vector corresponding to the known chronic disease quality control instance to obtain the complementary state vector corresponding to the known chronic disease quality control instance; Determine the first inference confidence of the known chronic disease quality control instance compared with the remaining chronic disease quality control states except the target known chronic disease quality control state according to the first confidence and the complementary state vector corresponding to the known chronic disease quality control instance.

4. The method according to claim 1, wherein The determining the first simulation error of the known chronic disease quality control instance compared with the corresponding target simulated chronic disease quality control state according to the first inference confidence includes: Obtain the target simulated chronic disease quality control state corresponding to the known chronic disease quality control instance; Determine the first simulation error of the known chronic disease quality control instance compared with the target simulated chronic disease quality control state according to the first inference confidence and the target value of the simulated chronic disease quality control state corresponding to the target simulated chronic disease quality control state.

5. The method according to claim 4, wherein The obtaining the target simulated chronic disease quality control state corresponding to the known chronic disease quality control instance includes: Determine the first key features corresponding to each simulated chronic disease quality control state in the feature domain, and the first key features are the key features corresponding to each simulated chronic disease quality control state in this round of training; Determine the similarity between the feature vector of the known chronic disease quality control instance and each first key feature respectively; Map the simulated chronic disease quality control state corresponding to the highest similarity value among multiple said similarities to the target simulated chronic disease quality control state corresponding to the known chronic disease quality control instance.

6. The method according to claim 1, characterized in that, The using the initial recognition model to infer the simulated chronic disease quality control instance and determining the second error of the simulated chronic disease quality control instance compared with the second target chronic disease quality control state according to the second inference value generated by the inference includes: Using the initial recognition model to extract the feature vector of the simulated chronic disease quality control instance and making an inference based on the feature vector of the simulated chronic disease quality control instance to obtain the second confidence level of the simulated chronic disease quality control instance compared with all chronic disease quality control states; Determine at least one known chronic disease quality control instance used to construct the simulated chronic disease quality control instance, and determine the target simulated chronic disease quality control state corresponding to the simulated chronic disease quality control instance according to the target simulated chronic disease quality control state corresponding to the at least one known chronic disease quality control instance; Determine the second basic error of the simulated chronic disease quality control instance compared with the target simulated chronic disease quality control state according to the second confidence level and the target simulated chronic disease quality control state corresponding to the simulated chronic disease quality control instance; Obtain the state identification vector corresponding to the simulated chronic disease quality control instance according to the target known chronic disease quality control state corresponding to the simulated chronic disease quality control instance; Perform a logical complement generation operation on the state identification vector corresponding to the simulated chronic disease quality control instance to obtain the complementary state vector corresponding to the simulated chronic disease quality control instance; Determine the first inference confidence level of the known chronic disease quality control instance compared with the remaining chronic disease quality control states except the target known chronic disease quality control state according to the second confidence level and the complementary state vector corresponding to the simulated chronic disease quality control instance; Obtain the target known chronic disease quality control state corresponding to the simulated chronic disease quality control instance; Determine the second simulated error of the simulated chronic disease quality control instance compared with the target known chronic disease quality control state according to the second inference confidence level and the known chronic disease quality control state target value corresponding to the target known chronic disease quality control state; Determine the second error of the simulated chronic disease quality control instance compared with the second target chronic disease quality control state according to the second basic error and the second simulated error.

7. The method according to claim 6, wherein The obtaining the target known chronic disease quality control state corresponding to the simulated chronic disease quality control instance includes: Determine the second key feature corresponding to each known chronic disease quality control state in the feature domain, and the second key feature is the key feature corresponding to the known chronic disease quality control state in this round of training; Respectively determine the similarity between the feature vector of the simulated chronic disease quality control instance and each second key feature; Map the known chronic disease quality control state corresponding to the second key feature with the highest similarity value among each similarity to the target known chronic disease quality control state corresponding to the simulated chronic disease quality control instance.

8. The method according to claim 1, wherein Obtaining a target simulated chronic disease quality control state corresponding to the new chronic disease quality control instance according to the deviation between the characterization vector and each target key feature includes: Determining the similarity between the characterization vector and each target key feature; Determining the simulated chronic disease quality control state mapped by the target key feature corresponding to the highest similarity value in each similarity as the target simulated chronic disease quality control state corresponding to the new chronic disease quality control instance.

9. A readable storage medium, characterized in that, The readable storage medium includes a computer program, and when the computer program runs, it controls the computer device where the readable storage medium is located to execute the method described in any one of claims 1-8.

Images