A prediction system for the therapeutic effect of traditional Chinese medicine prescriptions for acute cerebral infarction based on machine learning

Abstract

The invention discloses a prediction system for the therapeutic effect of Chinese medicine prescriptions for acute cerebral infarction based on machine learning, including: an acquisition module, which acquires the Chinese medicine prescriptions for acute cerebral infarction to be predicted; a model generation module, which takes each Chinese medicine in the Chinese medicine prescriptions for acute cerebral infarction to be predicted as a node, establishes undirected edges between all the Chinese medicines contained in the current prescription, and generates a complex network model; a proximity calculation module, which calculates the relative proximity of each node based on the complex network model; a feature selection module, which sorts the relative proximity of all nodes in the Chinese medicine prescriptions for acute cerebral infarction to be predicted in descending order, and selects several relative proximity degrees with the highest sorting as the features of the current Chinese medicine prescription sample; a prediction module, which obtains the prediction result of the therapeutic effect of the Chinese medicine prescription based on the features of the current Chinese medicine prescription sample and the trained machine learning model. Through the trained random forest model, the drug efficacy prediction of acute cerebral infarction disease is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of drug efficacy prediction, and in particular to a Chinese medicine prescription treatment effect prediction system for acute cerebral infarction based on machine learning. Background Art

[0002] The statements in this section merely mention background art related to the present invention and do not necessarily constitute prior art.

[0003] At present, the research and development of drugs for acute cerebral infarction faces problems such as insufficient efficacy and incomplete evaluation of adverse drug reactions. Therefore, we continue to develop new drug treatment effect prediction systems for acute cerebral infarction.

[0004] The inventors have found that existing drug efficacy screening methods often do not take into account Chinese herbal medicine prescriptions, or do not take into account which Chinese herbal medicines in the Chinese herbal medicine prescriptions are the key drugs that affect the therapeutic effect; as a result, the drug efficacy prediction accuracy of existing drug efficacy analysis systems is not accurate enough.

[0005] Regarding the Chinese invention patent, authorization announcement number CN109411033B, patent name: A drug efficacy screening method based on a complex network. The inventor believes that although it solves the screening of drug efficacy, it does not solve the screening of drug efficacy for diseases such as acute cerebral infarction, let alone the screening of drug efficacy of traditional Chinese medicine prescriptions for diseases such as acute cerebral infarction. Summary of the invention

[0006] In order to solve the deficiencies of the prior art, the present invention provides a prediction system for the therapeutic effect of Chinese medicine prescriptions for acute cerebral infarction based on machine learning;

[0007] In a first aspect, the present invention provides a prediction system for the therapeutic effect of Chinese medicine prescriptions for acute cerebral infarction based on machine learning;

[0008] The prediction system of the therapeutic effect of traditional Chinese medicine prescriptions for acute cerebral infarction based on machine learning includes:

[0009] An acquisition module is configured to: acquire a Chinese medicine prescription for acute cerebral infarction to be predicted;

[0010] A model generation module is configured to: take each Chinese medicine in the Chinese medicine prescription for acute cerebral infarction to be predicted as a node, establish undirected edges between all the Chinese medicines contained in the current prescription, and generate a complex network model;

[0011] A proximity calculation module, configured to: calculate the relative proximity of each node based on the complex network model;

[0012] A feature selection module is configured to: sort the relative closeness of all nodes in the predicted acute cerebral infarction Chinese medicine prescription in descending order, and select several relative closenesses with the highest order as the features of the current Chinese medicine prescription sample;

[0013] The prediction module is configured to obtain a prediction result of the therapeutic effect of the current Chinese medicine prescription based on the characteristics of the current Chinese medicine prescription sample and the trained machine learning model.

[0014] In a second aspect, the present invention further provides an electronic device, comprising:

[0015] a memory for non-transitory storage of computer-readable instructions; and

[0016] a processor for executing the computer readable instructions,

[0017] When the computer readable instructions are executed by the processor, the following steps are performed:

[0018] Obtain the Chinese medicine prescription for acute cerebral infarction to be predicted;

[0019] Taking each Chinese medicine in the Chinese medicine prescription for acute cerebral infarction to be predicted as a node, establishing undirected edges between all the Chinese medicines contained in the current prescription, and generating a complex network model;

[0020] Based on the complex network model, calculating the relative proximity of each node;

[0021] The relative closeness of all nodes in the predicted Chinese medicine prescription for acute cerebral infarction is sorted in descending order, and several relative closenesses with the highest order are selected as the features of the current Chinese medicine prescription sample;

[0022] Based on the characteristics of the current Chinese medicine prescription samples and the trained machine learning model, the prediction results of the therapeutic effect of the current Chinese medicine prescription are obtained.

[0023] In a third aspect, the present invention further provides a storage medium that non-temporarily stores computer-readable instructions, wherein when the non-temporary computer-readable instructions are executed by a computer, the following steps are performed:

[0024] Obtain the Chinese medicine prescription for acute cerebral infarction to be predicted;

[0025] Taking each Chinese medicine in the Chinese medicine prescription for acute cerebral infarction to be predicted as a node, establishing undirected edges between all the Chinese medicines contained in the current prescription, and generating a complex network model;

[0026] Based on the complex network model, calculating the relative proximity of each node;

[0027] The relative closeness of all nodes in the predicted Chinese medicine prescription for acute cerebral infarction is sorted in descending order, and several relative closenesses with the highest order are selected as the features of the current Chinese medicine prescription sample;

[0028] Based on the characteristics of the current Chinese medicine prescription samples and the trained machine learning model, the prediction results of the therapeutic effect of the current Chinese medicine prescription are obtained.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. The present invention uses a trained random forest model to predict the efficacy of drugs for diseases such as acute cerebral infarction;

[0031] 2. The present invention constructs a complex network model of the same traditional Chinese medicine document and calculates the node attribute matrix of each node in the complex network model to obtain the relative proximity of each node, thereby solving the screening of the most important several traditional Chinese medicines in each traditional Chinese medicine prescription and laying the foundation for the high-precision training of the random forest model.

[0032] 3. It can greatly reduce the parameters required for model training and effectively predict the six outcome indicators of the total effective rate of the prescription, BI score improvement, FIB level improvement, PV improvement, HSV level improvement and Lp-PLA2 level improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0034] Figure 1 This is a system function module diagram of embodiment 1. DETAILED DESCRIPTION

[0035] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0036] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0037] In this embodiment, all data is obtained in compliance with laws and regulations and based on the user's consent, and is used legally.

[0038] Complex network refers to a network with some or all of the properties of self-organization, self-similarity, attractor, small world, and scale-free. In our real life, many complex systems can be modeled as a complex network for analysis. Complex network is not only a form of data representation, but also a means of scientific research. Research on complex network has received extensive attention and research.

[0039] Random Forest is a flexible and easy-to-use machine learning algorithm that gives good results in most cases even without hyperparameter tuning. It can be used for classification and regression tasks.

[0040] The relative proximity of each drug node is calculated through a complex network to generate training and test samples, and the random forest model is trained.

[0041] Embodiment 1

[0042] This embodiment provides a prediction system for the therapeutic effect of Chinese medicine prescriptions for acute cerebral infarction based on machine learning;

[0043] like Figure 1 As shown, the prediction system of the therapeutic effect of Chinese medicine prescriptions for acute cerebral infarction based on machine learning includes:

[0044] An acquisition module is configured to: acquire a Chinese medicine prescription for acute cerebral infarction to be predicted;

[0045] A model generation module is configured to: take each Chinese medicine in the Chinese medicine prescription for acute cerebral infarction to be predicted as a node, establish undirected edges between all the Chinese medicines contained in the current prescription, and generate a complex network model;

[0046] A proximity calculation module, configured to: calculate the relative proximity of each node based on the complex network model;

[0047] A feature selection module is configured to: sort the relative closeness of all nodes in the predicted acute cerebral infarction Chinese medicine prescription in descending order, and select several relative closenesses with the highest order as the features of the current Chinese medicine prescription sample;

[0048] The prediction module is configured to obtain a prediction result of the therapeutic effect of the current Chinese medicine prescription based on the characteristics of the current Chinese medicine prescription sample and the trained machine learning model.

[0049] Furthermore, a feature preprocessing module is provided between the feature selection module and the prediction module;

[0050] The feature preprocessing module is configured to: perform normalization and dimensionality reduction processing on the features of the selected current Chinese medicine prescription sample.

[0051] Furthermore, the acquisition module: acquires a Chinese medicine prescription for acute cerebral infarction to be predicted; wherein a Chinese medicine prescription includes several Chinese medicine names.

[0052] Furthermore, the model generation module: takes each Chinese medicine in the Chinese medicine prescription for acute cerebral infarction to be predicted as a node, establishes undirected edges between all the Chinese medicines contained in the current prescription, and generates an undirected and unweighted complex network model; specifically includes:

[0053] Each Chinese medicine in the Chinese medicine prescription for acute cerebral infarction to be predicted is regarded as a node, and undirected edges are established between all the Chinese medicines contained in the current prescription to generate a complex network model.

[0054] Furthermore, the proximity calculation module calculates the relative proximity of each node based on the complex network model; the specific steps include:

[0055] Based on the complex network model, several attributes of each node are calculated; the attributes of each node include: the degree of each node, the betweenness centrality of each node, the closeness centrality of each node and the adjacency information entropy of each node;

[0056] Based on all the attributes of each node, the relative closeness of each node is calculated.

[0057] The degree of each node is calculated by traversing the complex network adjacency matrix and adding up the numbers in each row to get the node degree corresponding to this row.

[0058] Among them, the betweenness centrality of each node is calculated as follows:

[0059]

[0060] Among them, u is the node to be calculated; p is the total number of shortest paths between node s and node t; p(u) is the number of shortest paths between s and t passing through node u.

[0061] The closeness centrality of each node is calculated as follows: on the basis of calculating the shortest paths between all node pairs, the sum of the distances from each node to other nodes is calculated, and then the inverse of the obtained sum is taken to determine the closeness centrality of each node.

[0062] The adjacency information entropy of each node is calculated as follows: Calculate the sum of the neighbor node degrees of node i and node j of node i: j ; Divide the degree of node i by A jGet the probability P of a node being selected by its neighboring nodes ij ; Calculate the adjacent information entropy E of node i i :

[0063] E i =-∑(P ij log2P ij ).

[0064] The relative closeness of each node is calculated, including:

[0065] The node degree d, the node betweenness centrality bc, the node closeness centrality cc and the node adjacency information entropy ae are used as the four columns of the decision matrix. The decision matrix D = {x ij} m*n The form is as follows:

[0066]

[0067] The decision matrix element x ij Normalized to r ij :

[0068]

[0069] Normalize each element of the decision matrix, and then set a 1 / 4 weight for each element after normalization;

[0070] Then, the decision matrix element v after setting the weight ij To standardize:

[0071]

[0072] The standardized weighted decision matrix is ​​obtained.

[0073] Based on the standardized weighted decision matrix, the relative closeness of each node is calculated; specifically, it includes:

[0074] Take the maximum value of each column of the standardized weighted decision matrix as the optimal solution, and add the optimal solution to the optimal solution set;

[0075] Take the minimum value of each column of the standardized weighted decision matrix as the worst solution, and add the worst solution to the worst solution set;

[0076] The relative entropy of the node is calculated using the optimal solution set, the worst solution set, and each row of the standardized weighted decision matrix;

[0077] Based on the relative entropy of each node, the relative closeness of each node is calculated.

[0078] Among them, the maximum value of each column of the standardized weighted decision matrix is ​​taken as the optimal solution, and the optimal solution is added to the optimal solution set; specifically including:

[0079]

[0080] Among them, A + represents the optimal solution set, T1 and T2 represent the welfare standard and cost standard respectively, and v is the element in T1 and T2. i is A + The serial numbers of the four elements in correspond to the four columns of the weighted decision matrix.

[0081] Among them, the minimum value of each column of the standardized weighted decision matrix is ​​taken as the worst solution, and the worst solution is added to the worst solution set; specifically including:

[0082]

[0083] Among them, A - represents the worst solution set, T1 and T2 represent the welfare standard and cost standard respectively, and v is the element in T1 and T2. i is A - The serial numbers of the four elements in correspond to the four columns of the weighted decision matrix.

[0084] Among them, the relative entropy of the node is calculated using the optimal solution set, the worst solution set and each row of data in the standardized weighted decision matrix; specifically including:

[0085]

[0086]

[0087] in, represents the relative entropy corresponding to the optimal solution, represents the relative entropy corresponding to the worst solution, represents the elements in the optimal solution set, represents the element in the worst solution set, v i Represents the element corresponding to the node in the decision matrix.

[0088] Based on the relative entropy of each node, the relative closeness of each node is calculated; specifically, it includes:

[0089]

[0090] Among them, C i represents the relative proximity of each node, represents the relative entropy corresponding to the optimal solution, Represents the relative entropy corresponding to the worst solution.

[0091] Furthermore, the feature selection module: sorts the relative closeness of all nodes in the predicted acute cerebral infarction Chinese medicine prescription in descending order, and selects several relative closenesses with the highest order as the features of the current Chinese medicine prescription sample; specifically includes:

[0092] The relative closeness of all nodes in the predicted acute cerebral infarction Chinese medicine prescriptions are sorted in descending order, and the top ten relative closeness are selected as the characteristics of the current Chinese medicine prescription sample;

[0093] Among them, for sample features with less than ten relative closeness, the average value of the relative closeness of the current prescription is used to complete them.

[0094] Furthermore, the prediction module obtains the prediction result of the therapeutic effect of the current Chinese medicine prescription based on the characteristics of the current Chinese medicine prescription sample and the trained machine learning model; specifically includes:

[0095] The characteristics of the current Chinese medicine prescription samples are input into the trained machine learning model to obtain the prediction results of the therapeutic effect of the current Chinese medicine prescription.

[0096] Furthermore, the trained machine learning model is a trained random forest model.

[0097] Furthermore, the predicted results of the therapeutic effect of the current Chinese medicine prescription include six indicators: improvement in the total effective rate of treatment, improvement in the Barthel index (functional status of the patient's daily living activities) score, improvement in FIB (Fibrinogen, also known as fibrinogen) level, improvement in PV (polycythemia vera), improvement in HSV (Herpes simplexvirus) level and improvement in Lp-PLA2 (Lipoprotein-associated phospholipaseA2) level; those that meet the indicators are set to 1, and those that do not meet the indicators are set to 0.

[0098] Furthermore, the training process of the trained machine learning model includes:

[0099] Constructing a training set and a test set; both the training set and the test set are features of Chinese herbal prescriptions for acute cerebral infarction with known treatment effect labels; setting the ratio of the training set to the test set to 4:1;

[0100] Input the training set into the random forest model to train the model;

[0101] Input the test set into the random forest model, test the model, and obtain the trained machine learning model.

[0102] Furthermore, we first manually extracted more than 1,500 Chinese medicine prescriptions for the treatment of acute cerebral infarction and their corresponding outcome indicators. Using Chinese medicine as a node, undirected edges were established between Chinese medicines appearing in the same document to complete complex network modeling. The relative proximity of each Chinese medicine was calculated through the node attribute matrix to generate samples corresponding to each prescription. The samples were used to train the random forest model to predict the outcome indicators of the prescription.

[0103] In the process of constructing the training set, we manually extracted Chinese herbal prescriptions for acute cerebral infarction from the literature and their corresponding six outcome indicators, including improvement in total effective rate, improvement in BI score, improvement in FIB level, improvement in PV, improvement in HSV level and improvement in Lp-PLA2 level. The indicators that met the requirements were set to 1, otherwise they were set to 0.

[0104] Furthermore, the training set is input into the random forest model to train the model; specifically, the following steps are performed:

[0105] (1) Assume that the total number of training samples is N. We extract N samples with replacement, extracting 1 sample each time, and finally forming N samples. These N selected samples are used to train a decision tree and serve as the samples at the root node of the decision tree;

[0106] (2) Each sample has 10 attributes. When each node of the decision tree needs to be split, m attributes are randomly selected from these 10 attributes, satisfying the condition m<<10. Then, a certain strategy (information gain) is used to select one attribute from these m attributes as the splitting attribute of the node;

[0107] (3) During the decision tree formation process, each node must be split according to step (2) until it can no longer be split. No pruning is performed during the entire decision tree formation process;

[0108] (4) Establish a large number of decision trees according to steps (1) to (3), and determine the outcome label of the sample as 1 or 0 through voting results;

[0109] (5) Use the test set to verify the model accuracy and output the ROC curve.

[0110] The sample sources are detailed and reliable. The sample Chinese medicine prescriptions come from the systematic reviews and / or meta-analyses of Chinese medicine for the treatment of cerebral infarction published or updated from the self-built database of PubMed, Embase, Web of Science, and Cochrane Library to September 2020. The use of node relative proximity to generate samples greatly reduces the number of attributes contained in the samples and reduces the burden of model training. The model prediction accuracy and AUC are high.

[0111] The complex network contains 400 nodes, that is, 400 Chinese medicines. A common prescription sample is a vector of length 400. The position where the medicine appears in each prescription is set to 1, and the position where the medicine does not appear is set to 0, that is, the following form:

[0112] Table 1 Drug sample example table

[0113]

[0114]

[0115] Use the relative closeness C corresponding to the drugs contained in each prescription i Replace the medicine, and take the 10 largest relative closeness of each prescription to form a new sample of the following form:

[0116] Table 2 Relative closeness sample example table

[0117]

[0118] Process the samples and train a random forest model:

[0119] First, the column averages in the samples were used to supplement the relative closeness of the drugs in the prescriptions with less than 10 drugs, and the training set and the test set were divided into 4:1 ratio.

[0120] Let x represent the data in the sample, and x * Describe the normalized data. Use the following formula to normalize the sample data:

[0121]

[0122] The random forest model was trained after the sample data was reduced in dimension using the principal component analysis method.

[0123] Embodiment 2

[0124] This embodiment further provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, the one or more computer programs are stored in the memory, and when the electronic device is running, the processor executes the one or more computer programs stored in the memory, so that the electronic device performs the following steps:

[0125] Obtain the Chinese medicine prescription for acute cerebral infarction to be predicted;

[0126] Taking each Chinese medicine in the Chinese medicine prescription for acute cerebral infarction to be predicted as a node, establishing undirected edges between all the Chinese medicines contained in the current prescription, and generating a complex network model;

[0127] Based on the complex network model, calculating the relative proximity of each node;

[0128] The relative closeness of all nodes in the predicted Chinese medicine prescription for acute cerebral infarction is sorted in descending order, and several relative closenesses with the highest order are selected as the features of the current Chinese medicine prescription sample;

[0129] Based on the characteristics of the current Chinese medicine prescription samples and the trained machine learning model, the prediction results of the therapeutic effect of the current Chinese medicine prescription are obtained.

[0130] Embodiment 3

[0131] This embodiment further provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the following steps are performed:

[0132] Obtain the Chinese medicine prescription for acute cerebral infarction to be predicted;

[0133] Taking each Chinese medicine in the Chinese medicine prescription for acute cerebral infarction to be predicted as a node, establishing undirected edges between all the Chinese medicines contained in the current prescription, and generating a complex network model;

[0134] Based on the complex network model, calculating the relative proximity of each node;

[0135] The relative closeness of all nodes in the predicted Chinese medicine prescription for acute cerebral infarction is sorted in descending order, and several relative closenesses with the highest order are selected as the features of the current Chinese medicine prescription sample;

[0136] Based on the characteristics of the current Chinese medicine prescription samples and the trained machine learning model, the prediction results of the therapeutic effect of the current Chinese medicine prescription are obtained.

[0137] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A prediction system for the therapeutic effect of Chinese medicine prescriptions for acute cerebral infarction based on machine learning, characterized by: include: An acquisition module is configured to: acquire a Chinese medicine prescription for acute cerebral infarction to be predicted; A model generation module is configured to: take each Chinese medicine in the Chinese medicine prescription for acute cerebral infarction to be predicted as a node, establish undirected edges between all the Chinese medicines contained in the current prescription, and generate a complex network model; The proximity calculation module is configured to calculate the relative proximity of each node based on the complex network model. The specific steps include: Based on the complex network model, several attributes of each node are calculated; the attributes of each node include: the degree of each node, the betweenness centrality of each node, the closeness centrality of each node and the adjacency information entropy of each node; Based on all the attributes of each node, calculate the relative closeness of each node; Based on the degree of each node, the betweenness centrality of each node, the closeness centrality of each node, and the adjacency information entropy of each node, a standardized weighted decision matrix is ​​constructed; Take the maximum value of each column of the standardized weighted decision matrix as the optimal solution, and add the optimal solution to the optimal solution set; Take the minimum value of each column of the standardized weighted decision matrix as the worst solution, and add the worst solution to the worst solution set; The relative entropy of the node is calculated using the optimal solution set, the worst solution set, and each row of the standardized weighted decision matrix; Based on the relative entropy of each node, the relative closeness of each node is calculated; The degree of each node is calculated by traversing the complex network adjacency matrix and adding the numbers in each row to obtain the node degree corresponding to this row; The adjacency information entropy of each node is calculated by calculating the sum of the neighbor node degrees of neighbor node j of node i. ; Divide the degree of node i by Get the probability of a node being selected by its neighbor nodes ; Calculate the adjacent information entropy of node i : ； The calculation of the relative proximity of each node specifically includes: The degree of the node , the betweenness centrality of the node , the closeness centrality of the node and the node adjacency information entropy As the four columns of data in the decision matrix, the decision matrix The form is as follows: The decision matrix elements Normalized to : Normalize each element of the decision matrix, and then set a 1 / 4 weight for each element after normalization; Then, the decision matrix elements after setting the weights To standardize: Get the standardized weighted decision matrix; The feature selection module is configured to: sort the relative closeness of all nodes in the predicted acute cerebral infarction Chinese medicine prescription in descending order, and select several relative closenesses with the highest order as the features of the current Chinese medicine prescription sample; wherein, for sample features that are less than several relative closenesses, the average value of the relative closenesses of the current prescription is used to complete them; A prediction module, which is configured to: obtain a prediction result of the therapeutic effect of the current Chinese medicine prescription based on the characteristics of the current Chinese medicine prescription sample and the trained machine learning model; The trained machine learning model is a trained random forest model, and its training process includes: Constructing a training set and a test set; both the training set and the test set are features of Chinese herbal prescriptions for acute cerebral infarction with known treatment effect labels; setting the ratio of the training set and the test set; Input the training set into the random forest model to train the model; Input the test set into the random forest model, test the model, and obtain the trained machine learning model; In the process of constructing the training set, we manually extracted the Chinese herbal prescriptions for acute cerebral infarction from the literature and their corresponding six outcome indicators, namely, improvement in total effective rate, improvement in BI score, improvement in FIB level, improvement in PV, improvement in HSV level, and improvement in Lp-PLA2 level. The indicators that met the criteria were set to 1, otherwise they were set to 0. The training set is input into the random forest model to train the model; specifically, the following steps are included: (1) Assume that the total number of training samples is N. We extract N samples with replacement, extracting one sample each time, and finally forming N samples. We select the N good samples to train a decision tree and use them as the samples at the root node of the decision tree. (2) Each sample has 10 attributes. When each node of the decision tree needs to be split, m attributes are randomly selected from these 10 attributes, satisfying the condition m << 10; then a certain strategy, namely information gain, is used from these m attributes to select one attribute as the splitting attribute of the node; (3) During the decision tree formation process, each node must be split according to step (2) until it can no longer be split; no pruning is performed during the entire decision tree formation process; (4) Establish a large number of decision trees according to steps (1) to (3), and determine the outcome label of the sample as 1 or 0 through voting results; (5) Use the test set to verify the model accuracy and output the ROC curve.

2. The machine learning-based Chinese medicine prescription treatment effect prediction system for acute cerebral infarction according to claim 1, characterized in that a feature preprocessing module is further provided between the feature selection module and the prediction module; The feature preprocessing module is configured to: perform normalization and dimensionality reduction processing on the features of the selected current Chinese medicine prescription sample.

3. The machine learning-based Chinese medicine prescription treatment effect prediction system for acute cerebral infarction as claimed in claim 1, characterized in that: The method obtains the prediction result of the therapeutic effect of the current Chinese medicine prescription based on the characteristics of the current Chinese medicine prescription sample and the trained machine learning model; specifically includes: The characteristics of the current Chinese medicine prescription samples are input into the trained machine learning model to obtain the prediction results of the therapeutic effect of the current Chinese medicine prescription.

4. An electronic device, comprising: a memory for non-transitory storage of computer readable instructions; as well as a processor for executing the computer readable instructions, When the computer readable instructions are executed by the processor, the following steps are performed: Obtain the Chinese medicine prescription for acute cerebral infarction to be predicted; Taking each Chinese medicine in the Chinese medicine prescription for acute cerebral infarction to be predicted as a node, establishing undirected edges between all the Chinese medicines contained in the current prescription, and generating a complex network model; Based on the complex network model, the relative proximity of each node is calculated; the specific steps include: Based on the complex network model, several attributes of each node are calculated; the attributes of each node include: the degree of each node, the betweenness centrality of each node, the closeness centrality of each node and the adjacency information entropy of each node; Based on all the attributes of each node, calculate the relative closeness of each node; Based on the degree of each node, the betweenness centrality of each node, the closeness centrality of each node, and the adjacency information entropy of each node, a standardized weighted decision matrix is ​​constructed; Take the maximum value of each column of the standardized weighted decision matrix as the optimal solution, and add the optimal solution to the optimal solution set; Take the minimum value of each column of the standardized weighted decision matrix as the worst solution, and add the worst solution to the worst solution set; The relative entropy of the node is calculated using the optimal solution set, the worst solution set, and each row of the standardized weighted decision matrix; Based on the relative entropy of each node, the relative closeness of each node is calculated; The degree of each node is calculated by traversing the complex network adjacency matrix and adding the numbers in each row to obtain the node degree corresponding to this row; The adjacency information entropy of each node is calculated by calculating the sum of the neighbor node degrees of neighbor node j of node i. ; Divide the degree of node i by Get the probability of a node being selected by its neighbor nodes ; Calculate the adjacent information entropy of node i : ； The calculation of the relative proximity of each node specifically includes: The degree of the node , the betweenness centrality of the node , the closeness centrality of the node and the node adjacency information entropy As the four columns of data in the decision matrix, the decision matrix The form is as follows: The decision matrix elements Normalized to : Normalize each element of the decision matrix, and then set a 1 / 4 weight for each element after normalization; Then, the decision matrix elements after setting the weights To standardize: Get the standardized weighted decision matrix; The relative closeness of all nodes in the predicted Chinese medicine prescription for acute cerebral infarction is sorted in descending order, and several relative closenesses with the highest order are selected as the features of the current Chinese medicine prescription sample; among them, for sample features that are less than several relative closenesses, the average value of the relative closenesses of the current prescription is used to complete them; Based on the characteristics of the current Chinese medicine prescription samples and the trained machine learning model, the prediction results of the therapeutic effect of the current Chinese medicine prescription are obtained; The trained machine learning model is a trained random forest model, and its training process includes: Constructing a training set and a test set; both the training set and the test set are features of Chinese herbal prescriptions for acute cerebral infarction with known treatment effect labels; setting the ratio of the training set and the test set; Input the training set into the random forest model to train the model; Input the test set into the random forest model, test the model, and obtain the trained machine learning model; In the process of constructing the training set, we manually extracted the Chinese herbal prescriptions for acute cerebral infarction from the literature and their corresponding six outcome indicators, namely, improvement in total effective rate, improvement in BI score, improvement in FIB level, improvement in PV, improvement in HSV level, and improvement in Lp-PLA2 level. The indicators that met the criteria were set to 1, otherwise they were set to 0. The training set is input into the random forest model to train the model; specifically, the following steps are included: (1) Assume that the total number of training samples is N. We extract N samples with replacement, extracting one sample each time, and finally forming N samples. We select the N good samples to train a decision tree and use them as the samples at the root node of the decision tree. (2) Each sample has 10 attributes. When each node of the decision tree needs to be split, m attributes are randomly selected from these 10 attributes, satisfying the condition m << 10; then a certain strategy, namely information gain, is used from these m attributes to select one attribute as the splitting attribute of the node; (3) During the decision tree formation process, each node must be split according to step (2) until it can no longer be split; no pruning is performed during the entire decision tree formation process; (4) Establish a large number of decision trees according to steps (1) to (3), and determine the outcome label of the sample as 1 or 0 through voting results; (5) Use the test set to verify the model accuracy and output the ROC curve.

5. A storage medium characterized by being non-temporary wherein when the non-transitory computer-readable instructions are executed by a computer, the following steps are performed: Obtain the Chinese medicine prescription for acute cerebral infarction to be predicted; Taking each Chinese medicine in the Chinese medicine prescription for acute cerebral infarction to be predicted as a node, establishing undirected edges between all the Chinese medicines contained in the current prescription, and generating a complex network model; Based on the complex network model, the relative proximity of each node is calculated; the specific steps include: Based on the complex network model, several attributes of each node are calculated; the attributes of each node include: the degree of each node, the betweenness centrality of each node, the closeness centrality of each node and the adjacency information entropy of each node; Based on all the attributes of each node, calculate the relative closeness of each node; Based on the degree of each node, the betweenness centrality of each node, the closeness centrality of each node, and the adjacency information entropy of each node, a standardized weighted decision matrix is ​​constructed; Take the maximum value of each column of the standardized weighted decision matrix as the optimal solution, and add the optimal solution to the optimal solution set; Take the minimum value of each column of the standardized weighted decision matrix as the worst solution, and add the worst solution to the worst solution set; The relative entropy of the node is calculated using the optimal solution set, the worst solution set, and each row of the standardized weighted decision matrix; Based on the relative entropy of each node, the relative closeness of each node is calculated; The degree of each node is calculated by traversing the complex network adjacency matrix and adding the numbers in each row to obtain the node degree corresponding to this row; The adjacency information entropy of each node is calculated by calculating the sum of the neighbor node degrees of neighbor node j of node i. ; Divide the degree of node i by Get the probability of a node being selected by its neighbor nodes ; Calculate the adjacent information entropy of node i : ； The calculation of the relative proximity of each node specifically includes: The degree of the node , the betweenness centrality of the node , the closeness centrality of the node and the node adjacency information entropy As the four columns of data in the decision matrix, the decision matrix The form is as follows: The decision matrix elements Normalized to : Normalize each element of the decision matrix, and then set a 1 / 4 weight for each element after normalization; Then, the decision matrix elements after setting the weights To standardize: Get the standardized weighted decision matrix; The relative closeness of all nodes in the predicted Chinese medicine prescription for acute cerebral infarction is sorted in descending order, and several relative closenesses with the highest order are selected as the features of the current Chinese medicine prescription sample; among them, for sample features that are less than several relative closenesses, the average value of the relative closenesses of the current prescription is used to complete them; Based on the characteristics of the current Chinese medicine prescription samples and the trained machine learning model, the prediction results of the therapeutic effect of the current Chinese medicine prescription are obtained; The trained machine learning model is a trained random forest model, and its training process includes: Constructing a training set and a test set; both the training set and the test set are features of Chinese herbal prescriptions for acute cerebral infarction with known treatment effect labels; setting the ratio of the training set and the test set; Input the training set into the random forest model to train the model; Input the test set into the random forest model, test the model, and obtain the trained machine learning model; In the process of constructing the training set, we manually extracted the Chinese herbal prescriptions for acute cerebral infarction from the literature and their corresponding six outcome indicators, namely, improvement in total effective rate, improvement in BI score, improvement in FIB level, improvement in PV, improvement in HSV level, and improvement in Lp-PLA2 level. The indicators that met the criteria were set to 1, otherwise they were set to 0. The training set is input into the random forest model to train the model; specifically, the following steps are included: (1) Assume that the total number of training samples is N. We extract N samples with replacement, extracting one sample each time, and finally forming N samples. We select the N good samples to train a decision tree and use them as the samples at the root node of the decision tree. (2) Each sample has 10 attributes. When each node of the decision tree needs to be split, m attributes are randomly selected from these 10 attributes, satisfying the condition m << 10; then a certain strategy, namely information gain, is used from these m attributes to select one attribute as the splitting attribute of the node; (3) During the decision tree formation process, each node must be split according to step (2) until it can no longer be split; no pruning is performed during the entire decision tree formation process; (4) Establish a large number of decision trees according to steps (1) to (3), and determine the outcome label of the sample as 1 or 0 through voting results; (5) Use the test set to verify the model accuracy and output the ROC curve.

Images