Cardiovascular disease risk assessment system based on machine learning

Abstract

The invention discloses a cardiovascular disease risk assessment system based on machine learning. The cardiovascular disease risk assessment system comprises a data acquisition module, a data processing module, a cardiovascular disease risk assessment model building module, a model optimization module and a cardiovascular disease risk assessment module. The invention relates to the technical field of disease risk assessment, in particular to a cardiovascular disease risk assessment system based on machine learning, which improves the data quality by defining a depth distance function, a depth correlation degree, an evolution amplitude modulation function and an evolution weight; by defining a hybrid differential kernel and constructing a target and constraint conditions, the trend and speed of data change are captured, feature representation is enriched, and the accuracy and reliability of risk assessment are improved; by designing a double-strategy dynamic switching function and a local correction function, dynamic adjustment and intelligent switching of parameter search are realized, the search efficiency is improved, manual intervention and dependence are reduced, and the accuracy, reliability and stability of an optimization result are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of disease risk assessment, and in particular to a cardiovascular disease risk assessment system based on machine learning. Background Art

[0002] The cardiovascular disease risk assessment system based on machine learning is an innovative medical technology that uses machine learning algorithms to deeply mine and learn medical data to construct a cardiovascular disease risk assessment model, which can accurately predict and assess the risk of cardiovascular disease.

[0003] At the same time, traditional data processing methods have problems in cardiovascular disease risk assessment, such as inaccurate data division, lack of consideration of deep data correlation, unreasonable evolution of set center points, and susceptibility to interference from noisy data; in the construction of traditional cardiovascular disease risk assessment models, there are problems such as insufficient feature representation, weak model generalization ability, and low accuracy and reliability of risk assessment; in the optimization of traditional cardiovascular disease risk assessment models, there are problems such as low search efficiency, easy to fall into local optimality, and parameter adjustment relying on manual experience. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a cardiovascular disease risk assessment system based on machine learning. In view of the problems of inaccurate data division, lack of consideration of data deep correlation, unreasonable evolution of set center points and susceptibility to noise data interference in cardiovascular disease risk assessment in traditional data processing methods, this solution more comprehensively reflects the intrinsic relationship between data points by dividing neighborhoods, defining deep distance functions, calculating deep correlation, data division, designing evolutionary modulation functions, defining evolutionary weights, set center point evolution, iterative evolution and data cleaning, improves the ability to capture complex associations between data, and reduces the impact of noise data on the accuracy of cardiovascular disease risk assessment. In view of the problems of insufficient feature representation, weak model generalization ability and low accuracy and reliability of risk assessment in the construction of traditional cardiovascular disease risk assessment models, this solution adopts By defining mixed entropy, defining differential factors, designing mixed differential kernels, and constructing objectives and constraints, the trend and speed of data changes are captured, the feature representation is enriched, and the model can better handle complex patterns. While ensuring classification accuracy, it also has good generalization performance, thereby improving the accuracy and reliability of risk assessment. In response to the problems of low search efficiency, easy to fall into local optimality, and parameter adjustment relying on manual experience in the optimization of traditional cardiovascular disease risk assessment models, this scheme realizes dynamic adjustment and intelligent switching of parameter search by designing state switching factors, shrinkage factors, diffusion factors, dual-strategy dynamic switching functions, and local correction functions, guiding the search to approach the optimal solution, combining shrinkage and diffusion strategies, and intelligently selecting the search direction according to the current search state, thereby improving search efficiency, reducing manual intervention and dependence, and ensuring the accuracy, reliability, and stability of the optimization results.

[0005] The technical solution adopted by the present invention is as follows: the cardiovascular disease risk assessment system based on machine learning provided by the present invention includes a data acquisition module, a data processing module, a cardiovascular disease risk assessment model construction module, a model optimization module and a cardiovascular disease risk assessment module;

[0006] The data collection module collects age, gender, family medical history, living habit data, physiological index data and physical condition data of healthy users and users suffering from cardiovascular diseases;

[0007] The data processing module processes data by dividing neighborhoods, defining depth distance functions, calculating depth association, data partitioning, designing evolutionary modulation functions, defining evolutionary weights, set center point evolution, iterative evolution and data cleaning;

[0008] The cardiovascular disease risk assessment model construction module constructs a cardiovascular disease risk assessment model by defining mixed entropy, defining differential factors, designing mixed differential kernels, and constructing objectives and constraints;

[0009] The model optimization module optimizes the model by designing a state switching factor, a contraction factor, a diffusion factor, a dual-strategy dynamic switching function and a local correction function;

[0010] The cardiovascular disease risk assessment module uses a cardiovascular disease risk assessment model to assess the user's cardiovascular disease risk level.

[0011] Furthermore, the data collection module collects the age, gender, family medical history, lifestyle data, physiological indicator data and physical condition data of healthy users and users with cardiovascular diseases; the lifestyle data includes smoking habits, drinking habits, exercise frequency and intensity; the physiological indicator data includes blood pressure, blood lipid level, blood sugar level and heart rate; the physical condition data includes three categories: healthy state, mild disease state and severe disease state.

[0012] Furthermore, the data processing module specifically includes the following contents:

[0013] Divide the neighborhood into Set as the neighborhood width, where Represents the total number of dimensions of the data points, Represents the Euclidean distance between the two data points that are farthest apart; sets the total number of data points within the neighborhood width of a data point as the neighborhood density of the data point;

[0014] Initialize the data set, calculate the neighborhood density of all data points, select the data point with the highest neighborhood density as the first set center point, remove the selected data point and the data points within the neighborhood width of the selected data point, recalculate the neighborhood density, select the data point with the highest neighborhood density again as the next set center point, and repeat the process until all data points are removed and all set centers are obtained;

[0015] Define the depth distance function as follows:

[0016] ;

[0017] Among them, x1 and x2 represent data points, represents the depth distance function, Represents the depth distance value between data point x1 and data point x2, represents the square root, d represents the index of the dimension of the data point, represents the standard deviation of the data point in the dth dimension, represents the linear rectification function, and Respectively represent the values ​​of data point x1 and data point x2 in the dth dimension, represents the depth distance adjustment factor, Represents the data value in the dth dimension The probability of occurrence, Represents the data value in the dth dimension The probability of occurrence, Represents the JS divergence between data point x1 and data point x2 in the dth dimension;

[0018] Calculate the depth correlation, expressed as follows:

[0019] ;

[0020] in, represents the depth correlation function, represents the deep correlation between data point x1 and data point x2, e represents a natural constant, represents the association weight, represents the cosine function, represents the angle between data point x1 and data point x2 in the dth dimension, represents the correlation adjustment factor, Indicates taking the absolute value;

[0021] Data partitioning: Calculate the deep correlation between all data points and each set center point, and assign the data points to the set to which the set center point with the highest deep correlation with the data points belongs;

[0022] Design the evolutionary modulation function, which is expressed as follows:

[0023] ;

[0024] Where i represents the index of the data point, represents the i-th data point, k represents the index of the center point of the set, represents the kth set center point before evolution, represents the evolutionary amplitude modulation function, represents the evolutionary amplitude modulation intensity between the i-th data point and the k-th set center point before evolution, and Respectively represent the standard deviation and mean of the depth distance between all data points and each set center point;

[0025] The evolutionary weight is defined as follows:

[0026] ;

[0027] in, represents the evolutionary weight between the i-th data point and the k-th set center point before evolution, j represents the index of the set center point, represents the j-th set center point before evolution, represents the deep correlation between the i-th data point and the k-th set center point before evolution, represents the deep correlation between the ith data point and the jth set center point before evolution, and K represents the total number of set centers;

[0028] The evolution of the central point of the set is expressed as follows:

[0029] ;

[0030] in, represents the kth set center point after evolution, represents the index of the data point in the kth set, represents the total number of data points in the kth set, Indicates data points, Indicates the Euclidean distance. Indicates taking the data point with the closest distance;

[0031] Iterative evolution, repeating data partitioning and set center point evolution until the set no longer changes;

[0032] Data cleaning: calculate the number of data points in each set, the mean number of data points in the set, the Euclidean distance between each set and the nearest set, and the mean Euclidean distance between all sets and the nearest set. When there is a set with less than 50% of the mean number of data points in the set and the Euclidean distance between this set and the nearest set is greater than the mean Euclidean distance between all sets and the nearest set, set this set is set as a noise set and all data points in this set are cleared.

[0033] Furthermore, the cardiovascular disease risk assessment model module is constructed, specifically including the following contents:

[0034] Set label data, set the body status data as the label data of the model. The three categories of body status data: healthy state, mild disease state and severe disease state correspond to the three levels of cardiovascular disease risk: low risk, medium risk and high risk;

[0035] The definition of mixing entropy is as follows:

[0036] ;

[0037] Among them, q and h represent the index of the data point, and represent the qth data point and the hth data point respectively, represents the mixing entropy between the qth data point and the hth data point, and Represent the values ​​of the d-th dimension of the q-th data point and the h-th data point, respectively. represents the variance of the data point in the dth dimension, represents the smoothing factor, represents the logarithmic function, represents the logarithmic adjustment parameter;

[0038] Define the difference factor as follows:

[0039] ;

[0040] in, represents the difference factor between the qth data point and the hth data point, represents the second-order difference of the qth data point, represents the second-order difference of the hth data point, Indicates the modulus length, represents the first-order difference of the qth data point, represents the first-order difference of the h-th data point, The modulus of the maximum value among the first-order differences of all data points;

[0041] Design the hybrid difference kernel, which is expressed as follows:

[0042] ;

[0043] in, represents the mixed difference kernel between the qth and hth data points, and denote the mixed weight and differential weight respectively;

[0044] Construct the objectives and constraints, expressed as follows:

[0045] ;

[0046] in, and denote the weight and bias of the hyperplane respectively, Represents finding the optimal weight and bias Make the objective function reach the minimum value, represents the regularization coefficient, represents the total number of data points, represents the relaxation parameter, represents the label of the qth data point, represents the Lagrange multiplier of the h-th data point.

[0047] Furthermore, the model optimization module specifically includes the following contents:

[0048] Initialization: set the parameters to be optimized, including regularization coefficient, relaxation parameter, mixing weight, differential weight and logarithmic adjustment parameter, create parameter search space, set the accuracy of the model as the fitness of the parameter point, set the maximum number of parameter searches, and randomly generate initial parameter points in the parameter search space;

[0049] The design state switching factor is expressed as follows:

[0050] ;

[0051] Where t represents the number of parameter searches. represents the state switching factor during the tth parameter search, T represents the maximum number of parameter searches, Represents a random number ranging from 0 to 1;

[0052] The shrinkage factor is defined as follows:

[0053] ;

[0054] in, represents the shrinkage factor during the t-th parameter search, It indicates the fitness of the parameter point with the highest fitness among all the parameter points at this time. It represents the fitness of a randomly selected parameter point at this time. Represents the mean fitness of all parameter points at this time, Indicates the fitness of the parameter point with the lowest fitness among all parameter points at this time;

[0055] The diffusion factor is defined as follows:

[0056] ;

[0057] in, represents the diffusion factor during the t-th parameter search, Indicates the position of the parameter point with the highest fitness among all parameter points at this time, Represents a random number between 0 and 2. represents the symbolic function, represents the average position of the generated initial parameter points, Represents a random number between 0 and 0.5. Represents the distance between the upper and lower bounds of the search space;

[0058] Design a dual-strategy dynamic switching function, expressed as follows:

[0059] ;

[0060] in, Indicates the position of the parameter point obtained by searching the dual-strategy dynamic switching function during the t+1th parameter search. represents the position of the parameter point obtained by the dual-strategy dynamic switching function search in the tth parameter search, Indicates the position of the parameter point with the lowest fitness among all parameter points at this time;

[0061] Design a local correction function, expressed as follows:

[0062] ;

[0063] in, Indicates the position of the parameter point obtained by searching the local correction function during the t+1th parameter search. represents the position of the parameter point obtained by searching the local correction function during the tth parameter search, It represents the fitness of the parameter point obtained by searching the local correction function during the t-th parameter search;

[0064] Dynamic search, set the fitness standard value, first perform parameter search of the dual strategy dynamic switching function, then perform parameter search of the local correction function, when the fitness of the parameter point is greater than the fitness standard value, The parameters of are set as model parameters and the search ends; when , search again; otherwise continue searching.

[0065] Furthermore, the cardiovascular disease risk assessment module collects the user's age, gender, family medical history, lifestyle data and physiological index data, inputs the data into a cardiovascular disease risk assessment model, and the model outputs the user's cardiovascular disease risk level.

[0066] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0067] (1) In view of the problems of inaccurate data division, lack of consideration of data depth association, unreasonable evolution of set center points and susceptibility to interference from noise data in cardiovascular disease risk assessment in traditional data processing methods, this scheme reflects the intrinsic relationship between data points more comprehensively by dividing the neighborhood, defining the depth distance function, calculating the depth association, data division, designing the evolutionary modulation function, defining the evolutionary weight, set center point evolution, iterative evolution and data cleaning, thereby improving the ability to capture complex associations between data and reducing the impact of noise data on the accuracy of cardiovascular disease risk assessment.

[0068] (2) In order to address the problems of insufficient feature representation, weak model generalization ability, and low accuracy and reliability of risk assessment in the construction of traditional cardiovascular disease risk assessment models, this scheme captures the trend and speed of data changes by defining mixed entropy, defining differential factors, designing mixed differential kernels, and constructing objectives and constraints, enriching the feature representation, and enabling the model to better handle complex patterns. It can also have good generalization performance while ensuring classification accuracy, thereby improving the accuracy and reliability of risk assessment.

[0069] (3) To address the problems of low search efficiency, easy to fall into local optimality, and parameter adjustment relying on manual experience in the optimization of traditional cardiovascular disease risk assessment models, this scheme realizes dynamic adjustment and intelligent switching of parameter search by designing state switching factors, shrinkage factors, diffusion factors, dual-strategy dynamic switching functions and local correction functions, guiding the search to approach the optimal solution. By combining shrinkage and diffusion strategies, the search direction is intelligently selected according to the current search state, which improves search efficiency, reduces manual intervention and dependence, and ensures the accuracy, reliability and stability of the optimization results. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A schematic diagram of a cardiovascular disease risk assessment system based on machine learning provided by the present invention;

[0071] Figure 2 is a schematic diagram of a data processing module;

[0072] Figure 3 A schematic diagram of the module for constructing a cardiovascular disease risk assessment model;

[0073] Figure 4 Schematic diagram of the model optimization module;

[0074] Figure 5 A schematic diagram of a dynamic search;

[0075] Figure 6 It is a visualization diagram of the collection in the data processing module;

[0076] Figure 7 This is a diagram showing the effect of the model training process;

[0077] Figure 8 This is a diagram showing the effect of the model optimization process.

[0078] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0079] The technical solutions 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 only part of the embodiments of the present invention, rather than all 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.

[0080] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0081] Example 1, see Figure 1 The cardiovascular disease risk assessment system based on machine learning provided by the present invention includes a data acquisition module, a data processing module, a cardiovascular disease risk assessment model construction module, a model optimization module and a cardiovascular disease risk assessment module;

[0082] The data collection module collects the age, gender, family history, lifestyle data, physiological index data and physical condition data of healthy users and users with cardiovascular diseases; the lifestyle data includes smoking habits, drinking conditions, exercise frequency and intensity; the physiological index data includes blood pressure, blood lipid level, blood sugar level and heart rate; the physical condition data includes three categories: healthy state, mild disease state and severe disease state; and sends the data to the data processing module;

[0083] The data processing module processes data by dividing neighborhoods, defining depth distance functions, calculating depth association, data partitioning, designing evolutionary modulation functions, defining evolutionary weights, set center point evolution, iterative evolution and data cleaning, and sends the data to a cardiovascular disease risk assessment model building module;

[0084] The cardiovascular disease risk assessment model building module builds a cardiovascular disease risk assessment model by defining mixed entropy, defining differential factors, designing mixed differential kernels, building objectives and constraints, and sending data to the model optimization module;

[0085] The model optimization module optimizes the model by designing a state switching factor, a contraction factor, a diffusion factor, a dual-strategy dynamic switching function, and a local correction function, and sends the data to the cardiovascular disease risk assessment module;

[0086] The cardiovascular disease risk assessment module uses a cardiovascular disease risk assessment model to assess the user's cardiovascular disease risk level.

[0087] Example 2, see Figure 1 and Figure 2 This embodiment is based on the above embodiment, and the data processing module specifically includes the following contents:

[0088] Divide the neighborhood into Set as the neighborhood width, where Represents the total number of dimensions of the data points, Represents the Euclidean distance between the two data points that are farthest apart; sets the total number of data points within the neighborhood width of a data point as the neighborhood density of the data point;

[0089] Initialize the data set, calculate the neighborhood density of all data points, select the data point with the highest neighborhood density as the first set center point, remove the selected data point and the data points within the neighborhood width of the selected data point, recalculate the neighborhood density, select the data point with the highest neighborhood density again as the next set center point, and repeat the process until all data points are removed and all set centers are obtained;

[0090] Define the depth distance function, introduce dimension normalization and JS divergence to measure the relationship between two data points, as shown below:

[0091] ;

[0092] Among them, x1 and x2 represent data points, represents the depth distance function, Represents the depth distance value between data point x1 and data point x2, represents the square root, d represents the index of the dimension of the data point, represents the standard deviation of the data point in the dth dimension, represents the linear rectification function, and Respectively represent the values ​​of data point x1 and data point x2 in the dth dimension, represents the depth distance adjustment factor, Represents the data value in the dth dimension The probability of occurrence, Represents the data value in the dth dimension The probability of occurrence, Represents the JS divergence between data point x1 and data point x2 in the dth dimension;

[0093] Calculate the depth correlation, fuse the depth distance and cosine similarity to evaluate the correlation between data points, as shown below:

[0094] ;

[0095] in, represents the depth correlation function, represents the deep correlation between data point x1 and data point x2, e represents a natural constant, represents the association weight, represents the cosine function, represents the angle between data point x1 and data point x2 in the dth dimension, represents the correlation adjustment factor, Indicates taking the absolute value;

[0096] Data partitioning: Calculate the deep correlation between all data points and each set center point, and assign the data points to the set to which the set center point with the highest deep correlation with the data points belongs;

[0097] The evolutionary modulation function is designed to realize the dynamic adjustment mechanism by introducing the depth distance and performing normalization, which is expressed as follows:

[0098] ;

[0099] Where i represents the index of the data point, represents the i-th data point, k represents the index of the center point of the set, represents the kth set center point before evolution, represents the evolutionary amplitude modulation function, represents the evolutionary amplitude modulation intensity between the i-th data point and the k-th set center point before evolution, and Respectively represent the standard deviation and mean of the depth distance between all data points and each set center point;

[0100] Define evolutionary weights, and assign weights based on the depth of association between data points and the center point of the set, as follows:

[0101] ;

[0102] in, represents the evolutionary weight between the i-th data point and the k-th set center point before evolution, j represents the index of the set center point, represents the j-th set center point before evolution, represents the deep correlation between the i-th data point and the k-th set center point before evolution, represents the deep correlation between the ith data point and the jth set center point before evolution, and K represents the total number of set centers;

[0103] The evolution of the collection center point combines the depth correlation and weight, adjusts the depth distance and the evolution amplitude modulation intensity, evaluates the relationship between the data point and the collection center point, and dynamically adjusts the collection center point, which is expressed as follows:

[0104] ;

[0105] in, represents the kth set center point after evolution, represents the index of the data point in the kth set, represents the total number of data points in the kth set, Indicates data points, Indicates the Euclidean distance. Indicates taking the data point with the closest distance;

[0106] Iterative evolution, repeating data partitioning and set center point evolution until the set no longer changes;

[0107] Data cleaning: calculate the number of data points in each set, the mean number of data points in the set, the Euclidean distance between each set and the nearest set, and the mean Euclidean distance between all sets and the nearest set. When there is a set with less than 50% of the mean number of data points in the set and the Euclidean distance between this set and the nearest set is greater than the mean Euclidean distance between all sets and the nearest set, set this set is set as a noise set and all data points in this set are cleared.

[0108] By performing the above operations, in order to address the problems of inaccurate data division, lack of consideration of data deep correlation, unreasonable evolution of set center points and susceptibility to interference from noise data in traditional data processing methods for cardiovascular disease risk assessment, this scheme divides the neighborhood, defines deep distance function, calculates deep correlation, divides data, designs evolutionary modulation function, defines evolutionary weights, evolves set center points, iterates evolution and cleans data, which more comprehensively reflects the intrinsic relationship between data points, improves the ability to capture complex correlations between data, and reduces the impact of noise data on the accuracy of cardiovascular disease risk assessment.

[0109] Example 3, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the cardiovascular disease risk assessment model module is constructed, specifically including the following contents:

[0110] Set label data, set the body status data as the label data of the model. The three categories of body status data: healthy state, mild disease state and severe disease state correspond to the three levels of cardiovascular disease risk: low risk, medium risk and high risk;

[0111] Define the mixed entropy by calculating the difference between two data points at the dimension level and introducing the dimension variance, which is expressed as follows:

[0112] ;

[0113] Among them, q and h represent the index of the data point, and represent the qth data point and the hth data point respectively, represents the mixing entropy between the qth data point and the hth data point, and Represent the values ​​of the d-th dimension of the q-th data point and the h-th data point, respectively. represents the variance of the data point in the dth dimension, represents the smoothing factor, represents the logarithmic function, represents the logarithmic adjustment parameter;

[0114] Define the difference factor, and capture the dynamic changes of data points in spatial position by introducing first-order difference and second-order difference, as shown below:

[0115] ;

[0116] in, represents the difference factor between the qth data point and the hth data point, represents the second-order difference of the qth data point, represents the second-order difference of the hth data point, Indicates the modulus length, represents the first-order difference of the qth data point, represents the first-order difference of the h-th data point, The modulus of the maximum value among the first-order differences of all data points;

[0117] Design a hybrid difference kernel, which is designed by fusing the hybrid entropy and the difference factor, as shown below:

[0118] ;

[0119] in, represents the mixed difference kernel between the qth and hth data points, and denote the mixed weight and differential weight respectively;

[0120] Construct the objectives and constraints, expressed as follows:

[0121] ;

[0122] in, and denote the weight and bias of the hyperplane respectively, Represents finding the optimal weight and bias Make the objective function reach the minimum value, represents the regularization coefficient, represents the total number of data points, represents the relaxation parameter, represents the label of the qth data point, represents the Lagrange multiplier of the h-th data point.

[0123] By performing the above operations, this solution aims to address the problems of insufficient feature representation, weak model generalization ability, and low accuracy and reliability of risk assessment in the construction of traditional cardiovascular disease risk assessment models. By defining mixed entropy, defining differential factors, designing mixed differential kernels, and constructing objectives and constraints, this solution captures the trend and speed of data changes, enriches the feature representation, and enables the model to better handle complex patterns. It can also have good generalization performance while ensuring classification accuracy, thereby improving the accuracy and reliability of risk assessment.

[0124] Example 4, see Figure 1 , Figure 4 and Figure 5 This embodiment is based on the above embodiment, and the model optimization module specifically includes the following contents:

[0125] Initialization: set the parameters to be optimized, including regularization coefficient, relaxation parameter, mixing weight, differential weight and logarithmic adjustment parameter, create parameter search space, set the accuracy of the model as the fitness of the parameter point, set the maximum number of parameter searches, and randomly generate initial parameter points in the parameter search space;

[0126] Design the state switching factor by considering the iteration of the search and introducing randomness, which is expressed as follows:

[0127] ;

[0128] Where t represents the number of parameter searches. represents the state switching factor during the tth parameter search, T represents the maximum number of parameter searches, Represents a random number ranging from 0 to 1;

[0129] Define the shrinkage factor and use the fitness difference to define the shrinkage factor, which is expressed as follows:

[0130] ;

[0131] in, represents the shrinkage factor during the t-th parameter search, It indicates the fitness of the parameter point with the highest fitness among all the parameter points at this time. It represents the fitness of a randomly selected parameter point at this time. Represents the mean fitness of all parameter points at this time, Indicates the fitness of the parameter point with the lowest fitness among all parameter points at this time;

[0132] Define the diffusion factor and perform diffusion search based on the average position of the initial parameter points, as follows:

[0133] ;

[0134] in, represents the diffusion factor during the t-th parameter search, Indicates the position of the parameter point with the highest fitness among all parameter points at this time, Represents a random number between 0 and 2. represents the symbolic function, represents the average position of the generated initial parameter points, Represents a random number between 0 and 0.5. Represents the distance between the upper and lower bounds of the search space;

[0135] Design a dual-strategy dynamic switching function, design a search strategy based on the contraction factor and the diffusion factor, and use the state switching factor to dynamically select the search strategy, as shown below:

[0136] ;

[0137] in, Indicates the position of the parameter point obtained by searching the dual-strategy dynamic switching function during the t+1th parameter search. represents the position of the parameter point obtained by the dual-strategy dynamic switching function search in the tth parameter search, Indicates the position of the parameter point with the lowest fitness among all parameter points at this time;

[0138] Design a local correction function, consider the direction of the parameter point with the highest fitness, and adjust the searched parameter point, as shown below:

[0139] ;

[0140] in, Indicates the position of the parameter point obtained by searching the local correction function during the t+1th parameter search. represents the position of the parameter point obtained by searching the local correction function during the tth parameter search, It represents the fitness of the parameter point obtained by searching the local correction function during the t-th parameter search;

[0141] Dynamic search, set the fitness standard value, first perform parameter search of the dual strategy dynamic switching function, then perform parameter search of the local correction function, when the fitness of the parameter point is greater than the fitness standard value, The parameters of are set as model parameters and the search ends; when , search again; otherwise continue searching.

[0142] By performing the above operations, in order to address the problems of low search efficiency, easy to fall into local optimality and parameter adjustment relying on manual experience in the optimization of traditional cardiovascular disease risk assessment models, this scheme realizes dynamic adjustment and intelligent switching of parameter search by designing state switching factors, contraction factors, diffusion factors, dual-strategy dynamic switching functions and local correction functions, guides the search to the optimal solution, combines contraction and diffusion strategies, and intelligently selects the search direction according to the current search state, thereby improving search efficiency, reducing manual intervention and dependence, and ensuring the accuracy, reliability and stability of the optimization results.

[0143] Example 5, see Figure 1 This embodiment is based on the above embodiment. The cardiovascular disease risk assessment module collects the user's age, gender, family medical history, lifestyle data and physiological indicator data, inputs the data into the cardiovascular disease risk assessment model, and the model outputs the user's cardiovascular disease risk level.

[0144] Example 6, see Figure 6 , Figure 7 and Figure 8 This embodiment is based on the above embodiment, and the specific implementation effects of the data processing module, the cardiovascular disease risk assessment model building module and the model optimization module are as follows:

[0145] The data processing module constructs a complete set of data points through iterative evolution, sets it as a noise set and removes all data points in the set;

[0146] Construct a cardiovascular disease risk assessment model module, train the cardiovascular disease risk assessment model, and obtain the loss and accuracy of the model training process. The training loss of the model gradually decreases with the increase of training times, and the accuracy of the model gradually increases with the increase of training times, and the overall performance of the model is improved;

[0147] The model optimization module uses a dual-strategy dynamic switching function and a local correction function to search for model parameters in the parameter space. As the number of parameter searches increases, the number of searched parameter points increases, and the parameter fitness changes. The parameter position with the highest fitness obtained from each parameter search is recorded. When the search is completed, the optimized parameters are obtained.

[0148] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0149] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0150] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A cardiovascular disease risk assessment system based on machine learning, characterized by: It includes data acquisition module, data processing module, cardiovascular disease risk assessment model building module, model optimization module and cardiovascular disease risk assessment module; The data collection module collects age, gender, family medical history, living habit data, physiological index data and physical condition data of healthy users and users suffering from cardiovascular diseases; The data processing module processes data by dividing neighborhoods, defining depth distance functions, calculating depth association, data partitioning, designing evolutionary modulation functions, defining evolutionary weights, set center point evolution, iterative evolution and data cleaning; Among them, the depth distance function is defined as follows: ； Among them, x1 and x2 represent data points, represents the depth distance function, Represents the depth distance value between data point x1 and data point x2, represents the square root, d represents the index of the dimension of the data point, Represents the total number of dimensions of the data points, represents the standard deviation of the data point in the dth dimension, represents the linear rectification function, and Respectively represent the values ​​of data point x1 and data point x2 in the dth dimension, represents the depth distance adjustment factor, Represents the data value in the dth dimension The probability of occurrence, Represents the data value in the dth dimension The probability of occurrence, Represents the JS divergence between data point x1 and data point x2 in the dth dimension; The cardiovascular disease risk assessment model construction module constructs a cardiovascular disease risk assessment model by defining mixed entropy, defining differential factors, designing mixed differential kernels, and constructing objectives and constraints; The model optimization module optimizes the model by designing a state switching factor, a contraction factor, a diffusion factor, a dual-strategy dynamic switching function and a local correction function; The cardiovascular disease risk assessment module uses a cardiovascular disease risk assessment model to assess the user's cardiovascular disease risk level.

2. The cardiovascular disease risk assessment system based on machine learning according to claim 1, characterized in that: The data processing module specifically includes the following contents: Divide the neighborhood into Set as the neighborhood width, where Represents the Euclidean distance between the two data points that are farthest apart; sets the total number of data points within the neighborhood width of a data point as the neighborhood density of the data point; Initialize the data set, calculate the neighborhood density of all data points, select the data point with the highest neighborhood density as the first set center point, remove the selected data point and the data points within the neighborhood width of the selected data point, recalculate the neighborhood density, select the data point with the highest neighborhood density again as the next set center point, and repeat the process until all data points are removed and all set centers are obtained; Define the depth distance function; Calculate the depth correlation, expressed as follows: ； in, represents the depth correlation function, represents the deep correlation between data point x1 and data point x2, e represents a natural constant, represents the association weight, represents the cosine function, represents the angle between data point x1 and data point x2 in the dth dimension, represents the correlation adjustment factor, Indicates taking the absolute value; Data partitioning: Calculate the depth of correlation between all data points and each set center point, and assign the data points to the set to which the set center point with the highest depth of correlation with the data points belongs; Design the evolutionary modulation function, which is expressed as follows: ； Where i represents the index of the data point, represents the i-th data point, k represents the index of the center point of the set, represents the kth set center point before evolution, represents the evolutionary amplitude modulation function, represents the evolutionary amplitude modulation intensity between the i-th data point and the k-th set center point before evolution, and Respectively represent the standard deviation and mean of the depth distance between all data points and each set center point; The evolutionary weight is defined as follows: ； in, represents the evolutionary weight between the i-th data point and the k-th set center point before evolution, j represents the index of the set center point, represents the j-th set center point before evolution, represents the deep correlation between the i-th data point and the k-th set center point before evolution, represents the deep correlation between the ith data point and the jth set center point before evolution, and K represents the total number of set centers; The evolution of the central point of the set is expressed as follows: ； in, represents the kth set center point after evolution, represents the index of the data point in the kth set, represents the total number of data points in the kth set, Indicates data points, Indicates the Euclidean distance. Indicates taking the data point with the closest distance; Iterative evolution, repeating data partitioning and set center point evolution until the set no longer changes; Data cleaning: calculate the number of data points in each set, the mean number of data points in the set, the Euclidean distance between each set and the nearest set, and the mean Euclidean distance between all sets and the nearest set. When there is a set with less than 50% of the mean number of data points in the set and the Euclidean distance between this set and the nearest set is greater than the mean Euclidean distance between all sets and the nearest set, set this set is set as a noise set and all data points in this set are cleared.

3. The cardiovascular disease risk assessment system based on machine learning according to claim 1, characterized in that: The cardiovascular disease risk assessment model module is constructed, specifically including the following contents: Set label data, set the body status data as the label data of the model. The three categories of body status data: healthy state, mild disease state and severe disease state correspond to the three levels of cardiovascular disease risk: low risk, medium risk and high risk; The mixed entropy is defined as follows: ； Among them, q and h represent the index of the data point, and represent the qth data point and the hth data point respectively, represents the mixing entropy between the qth data point and the hth data point, and Represent the values ​​of the d-th dimension of the q-th data point and the h-th data point, respectively. represents the variance of the data point in the dth dimension, represents the smoothing factor, represents the logarithmic function, represents the logarithmic adjustment parameter; Define the difference factor as follows: ； in, represents the difference factor between the qth data point and the hth data point, represents the second-order difference of the qth data point, represents the second-order difference of the hth data point, Indicates the modulus length, represents the first-order difference of the qth data point, represents the first-order difference of the h-th data point, The modulus of the maximum value among the first-order differences of all data points; Design the hybrid difference kernel, which is expressed as follows: ； in, represents the mixed difference kernel between the qth and hth data points, and denote the mixed weight and differential weight respectively; Construct the objectives and constraints, expressed as follows: ； in, and denote the weight and bias of the hyperplane respectively, Represents finding the optimal weight and bias Make the objective function reach the minimum value, represents the regularization coefficient, represents the total number of data points, represents the relaxation parameter, represents the label of the qth data point, represents the Lagrange multiplier of the h-th data point.

4. The cardiovascular disease risk assessment system based on machine learning according to claim 1, characterized in that: The model optimization module specifically includes the following contents: Initialization: set the parameters to be optimized, including regularization coefficient, relaxation parameter, mixing weight, differential weight and logarithmic adjustment parameter, create parameter search space, set the accuracy of the model as the fitness of the parameter point, set the maximum number of parameter searches, and randomly generate initial parameter points in the parameter search space; The design state switching factor is expressed as follows: ； Where t represents the number of parameter searches. represents the state switching factor during the tth parameter search, T represents the maximum number of parameter searches, Represents a random number ranging from 0 to 1; The shrinkage factor is defined as follows: ； in, represents the shrinkage factor during the t-th parameter search, It indicates the fitness of the parameter point with the highest fitness among all the parameter points at this time. It represents the fitness of a randomly selected parameter point at this time. Represents the mean fitness of all parameter points at this time, Indicates the fitness of the parameter point with the lowest fitness among all parameter points at this time; The diffusion factor is defined as follows: ； in, represents the diffusion factor during the t-th parameter search, Indicates the position of the parameter point with the highest fitness among all parameter points at this time, Represents a random number between 0 and 2. represents the symbolic function, represents the average position of the generated initial parameter points, Represents a random number between 0 and 0.

5. Represents the distance between the upper and lower bounds of the search space; Design a dual-strategy dynamic switching function, expressed as follows: ； in, Indicates the position of the parameter point obtained by searching the dual-strategy dynamic switching function during the t+1th parameter search. represents the position of the parameter point obtained by the dual-strategy dynamic switching function search in the tth parameter search, Indicates the position of the parameter point with the lowest fitness among all parameter points at this time; Design a local correction function, expressed as follows: ； in, represents the position of the parameter point obtained by searching the local correction function during the t+1th parameter search. represents the position of the parameter point obtained by searching the local correction function during the tth parameter search. It represents the fitness of the parameter point obtained by searching the local correction function during the t-th parameter search; Dynamic search, set the fitness standard value, first perform parameter search of the dual strategy dynamic switching function, then perform parameter search of the local correction function, when the fitness of the parameter point is greater than the fitness standard value, The parameters of are set as model parameters and the search ends; when , search again; otherwise continue searching.

5. The cardiovascular disease risk assessment system based on machine learning according to claim 1, characterized in that: The data collection module collects the age, gender, family medical history, lifestyle data, physiological index data and physical condition data of healthy users and users with cardiovascular diseases; the lifestyle data includes smoking habits, drinking conditions, exercise frequency and intensity; the physiological index data includes blood pressure, blood lipid level, blood sugar level and heart rate; the physical condition data includes three categories: healthy state, mild disease state and severe disease state; The cardiovascular disease risk assessment module collects the user's age, gender, family medical history, lifestyle data and physiological index data, inputs the data into a cardiovascular disease risk assessment model, and the model outputs the user's cardiovascular disease risk level.

Images