Traditional Chinese medicine curative effect evaluation system based on artificial intelligence
By defining dynamic neighborhoods and designing local density calculation functions in the Chinese medical efficiency evaluation system, building dynamic activation functions and model architectures, and optimizing parameter update rules, the flexibility and stability problems in data detection and model construction in traditional systems are solved in traditional systems. , achieving more accurate and efficient evaluation of traditional Chinese medicine efficacy.
Patent Information
- Application Number
- CN202411574925.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-06
- Publication Date
- 2024-12-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional Chinese medical efficacy evaluation system has problems such as insufficient flexibility of data detection methods, inaccurate density calculation, unreasonable clustering center initialization, inaccurate similarity measurement, unstable clustering process and poor abnormal data processing. Chinese medical efficacy evaluation model The problem of insufficient flexibility of activation function, single model structure, insufficient feature processing, incomplete perfection of loss function, and low parameter update efficiency and poor accuracy.
Improve the stability of data detection by defining dynamic neighborhood width, setting neighborhoods, designing local density calculation functions, initializing cluster centers, designing similarity calculation functions, data point allocation, cluster center update, cluster formation and abnormal data clearing to improve the stability of data detection Design dynamic scaling activation function, model architecture, hidden layer, residual connection, dynamic zeroing and loss functions to build a Chinese medical efficiency evaluation model; use parameter update initialization, parameter multi-directional regulatory factor, parameter multi-directional adaptation search and dynamic Auxiliary search function to optimize model parameters.
The accuracy and stability of data detection are improved, making the Chinese medical efficiency evaluation model more accurate, effective and comprehensive, improving the efficiency and accuracy of parameter updates, and optimizing the model performance and effect.
Smart Images

Figure CN119092056A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of efficacy evaluation, and specifically to a traditional Chinese medicine efficacy evaluation system based on artificial intelligence. Background Art
[0002] The TCM efficacy evaluation system based on artificial intelligence uses artificial intelligence technology and big data technology to construct a TCM efficacy evaluation model based on historical patient data during and after treatment, and evaluates the TCM efficacy of patients through the model. However, the traditional data detection method has problems such as inflexible neighborhood definition, inaccurate density calculation, unreasonable cluster center initialization, inaccurate similarity measurement, unstable clustering process, and poor abnormal data processing; the traditional TCM efficacy evaluation model has problems such as insufficient flexibility of activation function, single model structure, insufficient feature processing, and imperfect loss function; the parameter update method of the traditional TCM efficacy evaluation model has problems such as low efficiency and poor accuracy. Summary of the invention
[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a TCM efficacy evaluation system based on artificial intelligence. In view of the problems of inflexible neighborhood definition, inaccurate density calculation, unreasonable cluster center initialization, inaccurate similarity measurement, unstable clustering process and poor abnormal data processing in traditional data detection methods, this solution performs data detection by defining dynamic neighborhood width, setting neighborhood, designing local density calculation function, initializing cluster center, designing similarity calculation function, data point allocation, cluster center update, cluster formation, designing abnormality function and abnormal data removal, thereby improving the stability of the clustering process and making the data more accurate, reliable and effective; in view of the problems of insufficient flexibility of activation function, single model structure, insufficiently sophisticated feature processing and inaccurate loss function in traditional TCM efficacy evaluation model, the present invention provides a TCM efficacy evaluation system based on artificial intelligence. In view of the problems of inflexible neighborhood definition, inaccurate density calculation, unreasonable cluster center initialization, inaccurate similarity measurement, unstable clustering process and poor abnormal data processing in traditional data detection methods, the present invention performs data detection by defining dynamic neighborhood width, setting neighborhood, designing local density calculation function, initializing cluster center, designing similarity calculation function, data point allocation, cluster center update, cluster formation, designing abnormality function and removing abnormal data, thereby improving the stability of clustering process and making the data more accurate, reliable and effective; in view of the problems of insufficient flexibility of activation function, single model structure, insufficiently sophisticated feature processing and inaccurate loss function in traditional TCM efficacy evaluation model, the present invention provides a TCM efficacy evaluation system based on artificial intelligence. In order to solve the problem that the traditional TCM efficacy evaluation model is not perfect enough, this solution constructs a TCM efficacy evaluation model by designing dynamic scaling activation functions, designing model architecture, designing hidden layers, introducing residual connections, dynamically setting zeros, designing output layers, and designing loss functions, which improves the adaptability and expressiveness of the activation function and the model's ability to process complex information, making the TCM efficacy evaluation model more accurate, effective, and comprehensive. In view of the low efficiency and poor accuracy of the traditional TCM efficacy evaluation model parameter updating method, this solution updates the model parameters through parameter update initialization, setting parameter performance indicators, designing multi-directional parameter control factors, multi-directional parameter adaptive search, designing dynamic auxiliary search functions, and designing parameter update rules, which improves the efficiency, accuracy, and adaptability of parameter updating, and can better optimize model parameters and improve model performance and effects.
[0004] The technical solution adopted by the present invention is as follows: the TCM efficacy evaluation system based on artificial intelligence provided by the present invention includes a data acquisition module, a data detection module, a TCM efficacy evaluation model construction module, a model parameter update module and a TCM efficacy evaluation module;
[0005] The data collection module collects historical patient symptom data, tongue data, pulse data, facial diagnosis data and efficacy evaluation data during and after treatment; the tongue data includes the color, shape and cracks of the tongue coating; the pulse data includes the position, frequency and strength of the pulse; the facial diagnosis data includes the glossiness of the face, the expression of the facial features and the tightness of the facial skin; the efficacy evaluation data includes the data of recovery, effectiveness and ineffectiveness;
[0006] The data detection module performs data detection by defining dynamic neighborhood width, setting neighborhood, designing local density calculation function, initializing cluster center, designing similarity calculation function, data point allocation, cluster center update, cluster formation, designing abnormality function and removing abnormal data;
[0007] The module for constructing a TCM efficacy evaluation model constructs a TCM efficacy evaluation model by designing a dynamic scaling activation function, designing a model architecture, designing a hidden layer, introducing a residual connection, performing dynamic zeroing, designing an output layer, and designing a loss function;
[0008] The model parameter updating module updates the model parameters by initializing parameter updates, setting parameter performance indicators, designing multi-directional parameter control factors, multi-directional parameter adaptive search, designing dynamic auxiliary search functions, and designing parameter update rules;
[0009] The TCM efficacy evaluation module inputs the symptom data, tongue data, pulse data and facial diagnosis data of the patient to be evaluated into the TCM efficacy evaluation model, and the model processes and analyzes the patient's data and outputs the patient's efficacy evaluation result.
[0010] Furthermore, the data detection module performs data detection by defining dynamic neighborhood width, setting neighborhood, designing local density calculation function, initializing cluster center, designing similarity calculation function, data point allocation, cluster center update, cluster formation, designing abnormality function and abnormal data removal, which specifically includes the following contents:
[0011] Define the dynamic neighborhood width as follows:
[0012] ;
[0013] Where i represents the index of the data point, represents the neighborhood width of the i-th data point, Div represents the total number of features of the data, Represents the Euclidean distance length between the i-th data point and the nearest data point;
[0014] Set the neighborhood. For any data point, take the data point as the center, set the area within the neighborhood width as the neighborhood area, and set all data points in the neighborhood area as neighborhood data points.
[0015] Design the local density calculation function, which is expressed as follows:
[0016] ;
[0017] Where j represents the index of the data point, represents the local density of the ith data point, It means that the i-th data point is the center. The set of all data point indices within the width range of represents a symbol, e represents a natural constant, and Respectively represent the i-th data point and the j-th data point, dist(·) represents the Euclidean distance, represents the Euclidean distance between the i-th data point and the j-th data point, represents the smoothing factor, represents the distance correction weight, Represents the average distance from the i-th data point to all neighboring data points;
[0018] Initialize the cluster center, calculate the local density of all data points, and select the one with the highest local density data points as the initial cluster centers;
[0019] Design a similarity calculation function, expressed as follows:
[0020] ;
[0021] in, represents the similarity counting function, represents the similarity between the i-th data point and the j-th data point, d represents the feature index of the data point, represents the feature similarity weight, Represents the similarity standard value, and denote the d-th eigenvalue of the ith and j-th data points respectively, Indicates taking the absolute value, represents the local density of the jth data point, represents the attenuation speed adjustment factor;
[0022] Data point allocation: calculate the similarity between each data point and each cluster center, and assign the data point to the cluster to which the cluster center with the highest similarity belongs;
[0023] Cluster center update: calculate the average position of all data points in each cluster, and set the data point closest to the average position as the new cluster center;
[0024] Cluster formation, repeating data point allocation and cluster center update until the clustering process converges;
[0025] Design the abnormality function, which is expressed as follows:
[0026] ;
[0027] in, represents the abnormality of the i-th data point, represents the total number of neighborhood data points of the i-th data point, represents the distance between the ith data point and the center point of the cluster to which the ith data point belongs. Represents the distance between the jth data point and the center point of the cluster to which the jth data point belongs;
[0028] Abnormal data removal: Calculate the average data point abnormality of each cluster formed, and remove the data points in the cluster with the highest average data point abnormality.
[0029] Furthermore, the module for constructing a TCM efficacy evaluation model constructs a TCM efficacy evaluation model by designing a dynamic scaling activation function, designing a model architecture, designing a hidden layer, introducing a residual connection, performing dynamic zeroing, designing an output layer, and designing a loss function, specifically including the following contents:
[0030] Design a dynamic scaling activation function, expressed as follows:
[0031] ;
[0032] in, represents the input value of the dynamic scaling activation function, represents the dynamic scaling activation function, Indicates taking the absolute value, and represents the scale parameter, It means square root. represents the dynamic threshold, and represent the threshold position control parameter and the threshold width control parameter respectively;
[0033] Design the model architecture, which includes input layer, hidden layer, residual connection layer, dynamic zeroing layer and output layer;
[0034] Design the hidden layer, expressed as follows:
[0035] ;
[0036] in, represents the output value of the hidden layer, , and Represent the weight, input value and bias of the hidden layer respectively;
[0037] The residual connection is introduced, which is expressed as follows:
[0038] ;
[0039] in, Represents the output after the residual connection is introduced, represents the residual weight, , and Represent the weight, input value and bias of the residual connection layer respectively;
[0040] Perform dynamic zeroing, as shown below:
[0041] ;
[0042] Among them, u represents the index of the neuron in the dynamic zeroing layer, represents the output value of the u-th neuron in the dynamic zeroing layer, represents the input value of the u-th neuron in the dynamic zeroing layer, r1 represents a random number ranging from 0 to 1, represents the zeroing probability of the dynamic zeroing layer;
[0043] Design the output layer, expressed as follows:
[0044] ;
[0045] in, represents the output value of the output layer, softmax(·) represents the normalized exponential activation function, , and Represent the weight, input value and bias of the output layer respectively;
[0046] The loss function is designed as follows:
[0047] ;
[0048] in, represents the loss value of the model, M is the total number of data samples, p is the index of the data sample, and q represents the index of the model's TCM efficacy evaluation category. represents the total number of TCM efficacy evaluation categories, represents the category balance factor of TCM efficacy evaluation, represents the true label of the p-th data sample for the q-th TCM efficacy evaluation category, represents the probability that the pth data sample output by the model belongs to the qth category of TCM efficacy evaluation, log(·) represents the logarithm, represents the attenuation coefficient, h represents the layer index of the hidden layer, and H represents the total number of hidden layers. represents the total number of neurons in the hth hidden layer, represents the neuron weight matrix of the hth hidden layer, Represents the square of the L2 norm of the weight matrix.
[0049] Furthermore, the updating model parameter module updates the model parameters by initializing parameter updates, setting parameter performance indicators, designing parameter multi-directional control factors, multi-directional adaptive search of parameters, designing dynamic auxiliary search functions and designing parameter update rules, which specifically includes the following contents:
[0050] Initialize parameter update, determine the parameters for parameter update, including the scale parameters, threshold control parameters, weights and biases of the TCM efficacy evaluation model, create parameter update space, and randomly generate initial parameter update point clusters in the parameter update space;
[0051] Set the parameter performance index and set the inverse of the model's loss value as the parameter's performance value;
[0052] The multi-directional control factor of the design parameters is expressed as follows:
[0053] ;
[0054] Among them, t represents the number of current parameter updates, represents the multi-directional parameter control factor at the tth parameter update, T represents the maximum number of parameter updates, v represents the index of the parameter update point cluster, Represents the total number of parameter update points, represents the position of the vth parameter update point in the generated initial parameter update point cluster, Indicates the parameter point with the highest current global parameter performance value. represents the average position of all parameter points obtained after completing the tth parameter update, The sum of the changes in parameter performance of all parameter update points after completing the tth parameter update;
[0055] Parameter multi-directional adaptive search, expressed as follows:
[0056] ;
[0057] in, Indicates the position of the parameter point obtained by multi-directional adaptive search during the t+1th parameter update. Indicates the position of the parameter point obtained by multi-directional adaptive search during the tth parameter update. Represents a random number between 0 and 1. Indicates taking the absolute value, Indicates the position of the parameter point obtained by multi-directional adaptive search during the t-1th parameter update. It indicates the parameter point with the highest performance value among all the parameter points obtained after the tth parameter update. Indicates the parameter point with the lowest parameter performance value among all parameter points obtained after completing the tth parameter update;
[0058] Design a dynamic auxiliary search function, expressed as follows:
[0059] ;
[0060] in, Indicates the position of the parameter point obtained by dynamic auxiliary search during the t+1th parameter update, represents the performance value of the parameter point obtained by multi-directional adaptive search of the parameters at the vth parameter update point during the tth parameter update. represents the maximum parameter performance after the tth parameter update, It represents the average parameter performance of all parameter points obtained by multi-directional parameter adaptation search at the tth parameter update;
[0061] Design parameter update rules, set the maximum number of parameter updates, set the parameter performance threshold, calculate the parameter performance values corresponding to the initial parameter point cluster, perform multi-directional parameter adaptation search, calculate the parameter performance values corresponding to the searched parameter points, and if the parameter performance values corresponding to more than 50% of the parameter points have been improved, directly perform the next multi-directional parameter adaptation search, otherwise first perform a search for the dynamic auxiliary search function, and then perform the next multi-directional parameter adaptation search; if a parameter point with a parameter performance value greater than the parameter performance threshold appears during the parameter update process, stop the search and update the parameters of the parameter point with the highest current global parameter performance value as the model parameter; if the maximum number of parameter updates is reached, re-update the parameters; otherwise continue to iterate and update.
[0062] Furthermore, the TCM efficacy evaluation module inputs the symptom data, tongue data, pulse data and facial diagnosis data of the patient to be evaluated into the TCM efficacy evaluation model, and the model processes and analyzes the patient's data and outputs the patient's efficacy evaluation results.
[0063] The beneficial effects achieved by the present invention using the above scheme are as follows:
[0064] (1) In view of the problems of inflexible neighborhood definition, inaccurate density calculation, unreasonable cluster center initialization, inaccurate similarity measurement, unstable clustering process and poor abnormal data processing in traditional data detection methods, this scheme performs data detection by defining dynamic neighborhood width, setting neighborhood, designing local density calculation function, initializing cluster center, designing similarity calculation function, data point allocation, cluster center update, cluster formation, designing anomaly function and removing abnormal data, which improves the stability of the clustering process and makes the data more accurate, reliable and effective.
[0065] (2) In view of the problems of insufficient flexibility of activation functions, single model structure, insufficiently sophisticated feature processing and imperfect loss functions in traditional TCM efficacy evaluation models, this scheme constructs a TCM efficacy evaluation model by designing dynamic scaling activation functions, designing model architecture, designing hidden layers, introducing residual connections, performing dynamic zeroing, designing output layers and designing loss functions. This improves the adaptability and expression ability of activation functions, and improves the model's ability to process complex information, making the TCM efficacy evaluation model more accurate, effective and comprehensive.
[0066] (3) In view of the problems of low efficiency and poor accuracy in the parameter updating method of the traditional Chinese medicine efficacy evaluation model, this scheme updates the model parameters by initializing parameter updates, setting parameter performance indicators, designing multi-directional parameter control factors, multi-directional parameter adaptive search, designing dynamic auxiliary search functions and designing parameter update rules. This improves the efficiency, accuracy and adaptability of parameter updates, and can better optimize model parameters and improve model performance and effects. BRIEF DESCRIPTION OF THE DRAWINGS
[0067] Figure 1 A schematic diagram of a TCM efficacy evaluation system based on artificial intelligence provided by the present invention;
[0068] Figure 2 is a schematic diagram of a data detection module;
[0069] Figure 3 This is a schematic diagram of the module for constructing a TCM efficacy evaluation model;
[0070] Figure 4 A schematic diagram of a module for updating model parameters;
[0071] Figure 5Schematic diagram of the parameter updating process;
[0072] Figure 6 Schematic diagram of the model parameter performance detection process.
[0073] 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
[0074] 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.
[0075] 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.
[0076] Example 1, see Figure 1 The artificial intelligence-based TCM efficacy evaluation system provided by the present invention includes a data acquisition module, a data detection module, a TCM efficacy evaluation model construction module, a model parameter update module and a TCM efficacy evaluation module;
[0077] The data collection module collects historical patient symptom data, tongue data, pulse data, facial diagnosis data and efficacy evaluation data during and after treatment;
[0078] The data detection module performs data detection by defining dynamic neighborhood width, setting neighborhood, designing local density calculation function, initializing cluster center, designing similarity calculation function, data point allocation, cluster center update, cluster formation, designing abnormality function and removing abnormal data;
[0079] The module for constructing a TCM efficacy evaluation model constructs a TCM efficacy evaluation model by designing a dynamic scaling activation function, designing a model architecture, designing a hidden layer, introducing a residual connection, performing dynamic zeroing, designing an output layer, and designing a loss function;
[0080] The model parameter updating module updates the model parameters by initializing parameter updates, setting parameter performance indicators, designing multi-directional parameter control factors, multi-directional parameter adaptive search, designing dynamic auxiliary search functions, and designing parameter update rules;
[0081] The TCM efficacy evaluation module inputs the symptom data, tongue data, pulse data and facial diagnosis data of the patient to be evaluated into the TCM efficacy evaluation model, and the model outputs the patient's efficacy evaluation result.
[0082] Example 2, see Figure 1 This embodiment is based on the above embodiment, and the data acquisition module collects historical patient symptom data, tongue data, pulse data, facial diagnosis data and efficacy evaluation data during and after treatment; the tongue data includes the color, shape and cracks of the tongue coating; the pulse data includes the position, frequency and strength of the pulse; the facial diagnosis data includes the glossiness of the face, the expression of the facial features and the tightness of the facial skin; the efficacy evaluation data includes cure, effectiveness and ineffectiveness.
[0083] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment, and the data detection module specifically includes the following contents:
[0084] Define the dynamic neighborhood width as follows:
[0085] ;
[0086] Where i represents the index of the data point, represents the neighborhood width of the i-th data point, Div represents the total number of features of the data, Represents the Euclidean distance length between the i-th data point and the nearest data point;
[0087] Set the neighborhood. For any data point, take the data point as the center, set the area within the neighborhood width as the neighborhood area, and set all data points in the neighborhood area as neighborhood data points.
[0088] Design the local density calculation function, which is expressed as follows:
[0089] ;
[0090] Where j represents the index of the data point, represents the local density of the ith data point, It means that the i-th data point is the center. The set of all data point indices within the width range of represents a symbol, e represents a natural constant, and Respectively represent the i-th data point and the j-th data point, dist(·) represents the Euclidean distance, represents the Euclidean distance between the i-th data point and the j-th data point, represents the smoothing factor, represents the distance correction weight, Represents the average distance from the i-th data point to all neighboring data points;
[0091] Initialize the cluster center, calculate the local density of all data points, and select the one with the highest local density data points as the initial cluster centers;
[0092] Design a similarity calculation function, expressed as follows:
[0093] ;
[0094] in, represents the similarity counting function, represents the similarity between the i-th data point and the j-th data point, d represents the feature index of the data point, represents the feature similarity weight, Represents the similarity standard value, and denote the d-th eigenvalue of the ith and j-th data points respectively, Indicates taking the absolute value, represents the local density of the jth data point, represents the attenuation speed adjustment factor;
[0095] Data point allocation: calculate the similarity between each data point and each cluster center, and assign the data point to the cluster to which the cluster center with the highest similarity belongs;
[0096] Cluster center update: calculate the average position of all data points in each cluster, and set the data point closest to the average position as the new cluster center;
[0097] Cluster formation, repeating data point allocation and cluster center update until the clustering process converges;
[0098] Design the abnormality function, which is expressed as follows:
[0099] ;
[0100] in, represents the abnormality of the i-th data point, represents the total number of neighborhood data points of the i-th data point, represents the distance between the ith data point and the center point of the cluster to which the ith data point belongs. Represents the distance between the jth data point and the center point of the cluster to which the jth data point belongs;
[0101] Abnormal data removal: Calculate the average data point abnormality of each cluster formed, and remove the data points in the cluster with the highest average data point abnormality.
[0102] By performing the above operations, in order to address the problems of inflexible neighborhood definition, inaccurate density calculation, unreasonable cluster center initialization, inaccurate similarity measurement, unstable clustering process and poor abnormal data processing in traditional data detection methods, this solution performs data detection by defining dynamic neighborhood width, setting neighborhood, designing local density calculation function, initializing cluster center, designing similarity calculation function, data point allocation, cluster center update, cluster formation, designing anomaly function and removing abnormal data, thereby improving the stability of the clustering process and making the data more accurate, reliable and effective.
[0103] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the module for constructing a TCM efficacy evaluation model specifically includes the following contents:
[0104] Design a dynamic scaling activation function, expressed as follows:
[0105] ;
[0106] in, represents the input value of the dynamic scaling activation function, represents the dynamic scaling activation function, Indicates taking the absolute value, and represents the scale parameter, It means square root. represents the dynamic threshold, and represent the threshold position control parameter and the threshold width control parameter respectively;
[0107] Design the model architecture, which includes input layer, hidden layer, residual connection layer, dynamic zeroing layer and output layer;
[0108] Design the hidden layer, expressed as follows:
[0109] ;
[0110] in, represents the output value of the hidden layer, , and Represent the weight, input value and bias of the hidden layer respectively;
[0111] The residual connection is introduced, which is expressed as follows:
[0112] ;
[0113] in, Represents the output after the residual connection is introduced, represents the residual weight, , and Represent the weight, input value and bias of the residual connection layer respectively;
[0114] Perform dynamic zeroing, as shown below:
[0115] ;
[0116] Among them, u represents the index of the neuron in the dynamic zeroing layer, represents the output value of the u-th neuron in the dynamic zeroing layer, represents the input value of the u-th neuron in the dynamic zeroing layer, r1 represents a random number ranging from 0 to 1, represents the zeroing probability of the dynamic zeroing layer;
[0117] Design the output layer, expressed as follows:
[0118] ;
[0119] in, represents the output value of the output layer, softmax(·) represents the normalized exponential activation function, , and Represent the weight, input value and bias of the output layer respectively;
[0120] The loss function is designed as follows:
[0121] ;
[0122] in, represents the loss value of the model, M is the total number of data samples, p is the index of the data sample, and q represents the index of the model's TCM efficacy evaluation category. represents the total number of TCM efficacy evaluation categories, represents the category balance factor of TCM efficacy evaluation, represents the true label of the p-th data sample for the q-th TCM efficacy evaluation category, represents the probability that the pth data sample output by the model belongs to the qth category of TCM efficacy evaluation, log(·) represents the logarithm, represents the attenuation coefficient, h represents the layer index of the hidden layer, and H represents the total number of hidden layers. represents the total number of neurons in the hth hidden layer, represents the neuron weight matrix of the hth hidden layer, Represents the square of the L2 norm of the weight matrix.
[0123] By performing the above operations, in order to address the problems of insufficient flexibility of activation functions, single model structure, insufficient feature processing and imperfect loss function in the traditional TCM efficacy evaluation model, this solution constructs a TCM efficacy evaluation model by designing dynamic scaling activation functions, designing model architecture, designing hidden layers, introducing residual connections, performing dynamic zeroing, designing output layers, and designing loss functions. It improves the adaptability and expressiveness of the activation function, and improves the model's ability to process complex information, making the TCM efficacy evaluation model more accurate, effective, and comprehensive.
[0124] Example 5, see Figure 1 , Figure 4 , Figure 5 and Figure 6 This embodiment is based on the above embodiment, and the model parameter updating module specifically includes the following contents:
[0125] Initialize parameter update, determine the parameters for parameter update, including the scale parameters, threshold control parameters, weights and biases of the TCM efficacy evaluation model, create parameter update space, and randomly generate initial parameter update point clusters in the parameter update space;
[0126] Set the parameter performance index and set the inverse of the model's loss value as the parameter's performance value;
[0127] The multi-directional control factor of the design parameters is expressed as follows:
[0128] ;
[0129] Among them, t represents the number of current parameter updates, represents the multi-directional parameter control factor at the tth parameter update, T represents the maximum number of parameter updates, v represents the index of the parameter update point cluster, Represents the total number of parameter update points, represents the position of the vth parameter update point in the generated initial parameter update point cluster, Indicates the parameter point with the highest current global parameter performance value. represents the average position of all parameter points obtained after completing the tth parameter update, The sum of the changes in parameter performance of all parameter update points after completing the tth parameter update;
[0130] Parameter multi-directional adaptive search, expressed as follows:
[0131] ;
[0132] in, Indicates the position of the parameter point obtained by multi-directional adaptive search during the t+1th parameter update. Indicates the position of the parameter point obtained by multi-directional adaptive search during the tth parameter update. Represents a random number between 0 and 1. Indicates taking the absolute value, Indicates the position of the parameter point obtained by multi-directional adaptive search during the t-1th parameter update. It indicates the parameter point with the highest performance value among all the parameter points obtained after the tth parameter update. Indicates the parameter point with the lowest parameter performance value among all parameter points obtained after completing the tth parameter update;
[0133] Design a dynamic auxiliary search function, expressed as follows:
[0134] ;
[0135] in, Indicates the position of the parameter point obtained by dynamic auxiliary search during the t+1th parameter update, represents the performance value of the parameter point obtained by multi-directional adaptive search of the parameters at the vth parameter update point during the tth parameter update. represents the maximum parameter performance after the tth parameter update, It represents the average parameter performance of all parameter points obtained by multi-directional parameter adaptation search at the tth parameter update;
[0136] Design parameter update rules, set the maximum number of parameter updates, set the parameter performance threshold, calculate the parameter performance values corresponding to the initial parameter point cluster, perform multi-directional parameter adaptation search, calculate the parameter performance values corresponding to the searched parameter points, and if the parameter performance values corresponding to more than 50% of the parameter points have been improved, directly perform the next multi-directional parameter adaptation search, otherwise first perform a search for the dynamic auxiliary search function, and then perform the next multi-directional parameter adaptation search; if a parameter point with a parameter performance value greater than the parameter performance threshold appears during the parameter update process, stop the search and update the parameters of the parameter point with the highest current global parameter performance value as the model parameter; if the maximum number of parameter updates is reached, re-update the parameters; otherwise continue to iterate and update.
[0137] By performing the above operations, in order to address the problems of low efficiency and poor accuracy in the parameter updating method of the traditional Chinese medicine efficacy evaluation model, this solution updates the model parameters by initializing parameter updates, setting parameter performance indicators, designing multi-directional parameter control factors, multi-directional parameter adaptive search, designing dynamic auxiliary search functions and designing parameter update rules. This improves the efficiency, accuracy and adaptability of parameter updates, and can better optimize model parameters and improve model performance and effects.
[0138] 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.
[0139] 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.
[0140] 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. The TCM efficacy evaluation system based on artificial intelligence is characterized by: It includes data collection module, data detection module, TCM efficacy evaluation model building module, model parameter updating module and TCM efficacy evaluation module; The data collection module collects historical patient symptom data, tongue data, pulse data, facial diagnosis data and efficacy evaluation data during and after treatment; The data detection module performs data detection by defining dynamic neighborhood width, setting neighborhood, designing local density calculation function, initializing cluster center, designing similarity calculation function, data point allocation, cluster center update, cluster formation, designing abnormality function and removing abnormal data; The module for constructing a TCM efficacy evaluation model constructs a TCM efficacy evaluation model by designing a dynamic scaling activation function, designing a model architecture, designing a hidden layer, introducing a residual connection, performing dynamic zeroing, designing an output layer, and designing a loss function; The model parameter updating module updates the model parameters by initializing parameter updates, setting parameter performance indicators, designing multi-directional parameter control factors, multi-directional parameter adaptive search, designing dynamic auxiliary search functions, and designing parameter update rules; The TCM efficacy evaluation module inputs the symptom data, tongue data, pulse data and facial diagnosis data of the patient to be evaluated into the TCM efficacy evaluation model, and the model outputs the patient's efficacy evaluation result.
2. The artificial intelligence-based TCM efficacy evaluation system according to claim 1 is characterized by: The data detection module specifically includes the following contents: Define the dynamic neighborhood width as follows: ; Where i represents the index of the data point, represents the neighborhood width of the i-th data point, Div represents the total number of features of the data, Represents the Euclidean distance length between the i-th data point and the nearest data point; Set the neighborhood. For any data point, take the data point as the center, set the area within the neighborhood width as the neighborhood area, and set all data points in the neighborhood area as neighborhood data points. Design the local density calculation function, which is expressed as follows: ; Where j represents the index of the data point, represents the local density of the ith data point, It means that the i-th data point is the center. The set of all data point indices within the width range of represents a symbol, e represents a natural constant, and Respectively represent the i-th data point and the j-th data point, dist(·) represents the Euclidean distance, represents the Euclidean distance between the i-th data point and the j-th data point, represents the smoothing factor, represents the distance correction weight, Represents the average distance from the i-th data point to all neighboring data points; Initialize the cluster center, calculate the local density of all data points, and select the one with the highest local density data points as the initial cluster centers; Design a similarity calculation function, expressed as follows: ; in, represents the similarity counting function, represents the similarity between the i-th data point and the j-th data point, d represents the feature index of the data point, represents the feature similarity weight, Represents the similarity standard value, and denote the d-th eigenvalue of the ith and j-th data points respectively, Indicates taking the absolute value, represents the local density of the jth data point, It means square root. represents the attenuation speed adjustment factor; Data point allocation: calculate the similarity between each data point and each cluster center, and assign the data point to the cluster to which the cluster center with the highest similarity belongs; Cluster center update: calculate the average position of all data points in each cluster, and set the data point closest to the average position as the new cluster center; Cluster formation, repeating data point allocation and cluster center update until the clustering process converges; Design the abnormality function, which is expressed as follows: ; in, represents the abnormality of the i-th data point, represents the total number of neighborhood data points of the i-th data point, represents the distance between the ith data point and the center point of the cluster to which the ith data point belongs. Represents the distance between the jth data point and the center point of the cluster to which the jth data point belongs; Abnormal data removal: Calculate the average data point abnormality of each cluster formed, and remove the data points in the cluster with the highest average data point abnormality.
3. The artificial intelligence-based TCM efficacy evaluation system according to claim 1 is characterized in that: The module for constructing a TCM efficacy evaluation model specifically includes the following contents: Design a dynamic scaling activation function, expressed as follows: ; in, represents the input value of the dynamic scaling activation function, represents the dynamic scaling activation function, Indicates taking the absolute value, and represents the scale parameter, It means square root. represents the dynamic threshold, and represent the threshold position control parameter and the threshold width control parameter respectively; Design the model architecture, which includes input layer, hidden layer, residual connection layer, dynamic zeroing layer and output layer; Design the hidden layer, expressed as follows: ; in, represents the output value of the hidden layer, , and Represent the weight, input value and bias of the hidden layer respectively; The residual connection is introduced, which is expressed as follows: ; in, Represents the output after the residual connection is introduced, represents the residual weight, , and Represent the weight, input value and bias of the residual connection layer respectively; Perform dynamic zeroing, as shown below: ; Among them, u represents the index of the neuron in the dynamic zeroing layer, represents the output value of the u-th neuron in the dynamic zeroing layer, represents the input value of the u-th neuron in the dynamic zeroing layer, r1 represents a random number ranging from 0 to 1, represents the zeroing probability of the dynamic zeroing layer; Design the output layer, expressed as follows: ; in, represents the output value of the output layer, softmax(·) represents the normalized exponential activation function, , and Represent the weight, input value and bias of the output layer respectively; The loss function is designed as follows: ; in, represents the loss value of the model, M is the total number of data samples, p is the index of the data sample, and q represents the index of the model's TCM efficacy evaluation category. represents the total number of TCM efficacy evaluation categories, represents the category balance factor of TCM efficacy evaluation, represents the true label of the p-th data sample for the q-th TCM efficacy evaluation category, represents the probability that the pth data sample output by the model belongs to the qth category of TCM efficacy evaluation, log(·) represents the logarithm, represents the attenuation coefficient, h represents the layer index of the hidden layer, and H represents the total number of hidden layers. represents the total number of neurons in the hth hidden layer, represents the neuron weight matrix of the hth hidden layer, Represents the square of the L2 norm of the weight matrix.
4. The artificial intelligence-based TCM efficacy evaluation system according to claim 1 is characterized in that: The model parameter updating module specifically includes the following contents: Initialize parameter update, determine the parameters for parameter update, including the scale parameters, threshold control parameters, weights and biases of the TCM efficacy evaluation model, create parameter update space, and randomly generate initial parameter update point clusters in the parameter update space; Set the parameter performance index and set the inverse of the model's loss value as the parameter's performance value; The multi-directional control factor of the design parameters is expressed as follows: ; Among them, t represents the number of current parameter updates, represents the multi-directional parameter control factor at the tth parameter update, T represents the maximum number of parameter updates, v represents the index of the parameter update point cluster, Represents the total number of parameter update points, represents the position of the vth parameter update point in the generated initial parameter update point cluster, Indicates the parameter point with the highest current global parameter performance value. represents the average position of all parameter points obtained after completing the tth parameter update, The sum of the changes in parameter performance of all parameter update points after completing the tth parameter update; Parameter multi-directional adaptive search, expressed as follows: ; in, Indicates the position of the parameter point obtained by multi-directional adaptive search during the t+1th parameter update. Indicates the position of the parameter point obtained by multi-directional adaptive search during the tth parameter update. Represents a random number between 0 and 1. Indicates taking the absolute value, Indicates the position of the parameter point obtained by multi-directional adaptive search during the t-1th parameter update. It indicates the parameter point with the highest performance value among all the parameter points obtained after the tth parameter update. Indicates the parameter point with the lowest parameter performance value among all parameter points obtained after completing the tth parameter update; Design a dynamic auxiliary search function, expressed as follows: ; in, Indicates the position of the parameter point obtained by dynamic auxiliary search during the t+1th parameter update, represents the performance value of the parameter point obtained by multi-directional adaptive search of the parameters at the vth parameter update point during the tth parameter update. represents the maximum parameter performance after the tth parameter update, It represents the average parameter performance of all parameter points obtained by multi-directional parameter adaptation search at the tth parameter update; Design parameter update rules, set the maximum number of parameter updates, set the parameter performance threshold, calculate the parameter performance values corresponding to the initial parameter point cluster, perform multi-directional parameter adaptation search, calculate the parameter performance values corresponding to the searched parameter points, and if the parameter performance values corresponding to more than 50% of the parameter points have been improved, directly perform the next multi-directional parameter adaptation search, otherwise first perform a search for the dynamic auxiliary search function, and then perform the next multi-directional parameter adaptation search; if a parameter point with a parameter performance value greater than the parameter performance threshold appears during the parameter update process, stop the search and update the parameters of the parameter point with the highest current global parameter performance value as the model parameter; if the maximum number of parameter updates is reached, re-update the parameters; otherwise continue to iterate and update.
5. The artificial intelligence-based TCM efficacy evaluation system according to claim 1 is characterized by: The data acquisition module collects historical patient symptom data, tongue data, pulse data, facial diagnosis data and efficacy evaluation data during and after treatment; the tongue data includes the color, shape and cracks of the tongue coating; the pulse data includes the position, frequency and strength of the pulse; the facial diagnosis data includes the glossiness of the face, the expression of the facial features and the tightness of the facial skin; the efficacy evaluation data includes cure, effectiveness and ineffectiveness.
Citation Information
Patent Citations
Method for detecting abnormal target in hyperspectral remotely sensed image
CN101794437A
Weak light vehicle detection method based on multi-exposure generation fusion
CN116129375A
Lithium battery temperature real-time prediction model based on multi-time-space-width learning model and construction method
CN116258059A
Traditional Chinese medicine curative effect evaluation system and method for gout
CN116913475A
Systems and Methods for Implementing Smart Assistant Systems
US20240119932A1
Cited By
Gastrointestinal patient postoperative recovery management system based on big data
CN119786069A
Postoperative Recovery Management System for Gastrointestinal Patients Based on Big Data
CN119786069B
Rheumatism activity assessment system based on medical information processing
CN119993544A
Rheumatism Activity Assessment System Based on Medical Information Processing
CN119993544B
