A case quality and efficiency evaluation method and system based on PCA-BP-GA

Abstract

The present invention discloses a case quality and efficiency evaluation method and system based on PCA-BP-GA, the method comprising: S1, receiving case information to be evaluated, and organizing it into case-related data sets, the data sets including case quality and efficiency system indicator data of each court; S2, preprocessing the data set; using principal component analysis PCA to reduce the dimension of the data set, and extracting the main indicator data affecting the quality and efficiency of the case; S3, inputting the main indicator data affecting the quality and efficiency of the case extracted in step S2 into the case quality and efficiency evaluation model, and obtaining the quality and efficiency evaluation score of the corresponding case through the case quality and efficiency evaluation model; S4, completing the evaluation of the case according to the case quality and efficiency evaluation score. The present invention introduces principal component analysis and genetic algorithm into the BP neural network, so that the model can obtain better results and can scientifically and objectively evaluate the quality and efficiency of the case.

Description

Technical Field

[0001] The present invention relates to data mining technology, and in particular to a case quality and efficiency evaluation method and system based on PCA-BP-GA. Background Art

[0002] Information technology is constantly integrating with traditional industries, triggering far-reaching industrial changes and forming new production methods, industrial forms, business models, and so on. In recent years, the integration of information technology and court work has received more and more attention. Relying on artificial intelligence technologies such as intelligent voice and image recognition, iFLYTEK can form an intelligent recording system for the entire judicial process. Carry out intelligent transformation and promote the transformation of my country's judicial work to smart justice. Traditional case quality and efficiency evaluation often has drawbacks such as high subjectivity and inconsistent standards in various places. The deep integration of information technology and case quality and efficiency evaluation has become an important direction for my country's intelligent judicial transformation at present and in the future.

[0003] The purpose of establishing a scientific and objective quality and efficiency indicator system is to grasp the key points and essentials of trial quality and efficiency management, and then comprehensively improve the quality and efficiency of trials. By consulting the quality and efficiency evaluations of various places, conducting in-depth research on a large number of literature materials, consulting relevant legal personnel, and summarizing and summarizing the relevant data, a series of indicators related to quality and efficiency can be analyzed, which can evaluate the quality and efficiency of cases to a certain extent. After having a scientific quality and efficiency indicator system, if the evaluation and scoring of the quality and efficiency evaluation work is a more important task, whether the weight of each indicator is consistent, whether there is a difference in importance between the indicators, etc., these are all things that need to be considered. In the past, work was done based on human experience, which was subject to many difficulties such as high subjectivity, large workload, and difficult maintenance.

[0004] Using neural networks to model data has become the mainstream of various engineering project research. As a traditional feedforward neural network, BP network essentially realizes a mapping function from input to output, and has strong nonlinear mapping capabilities. And when training, BP neural network can automatically extract the "reasonable rules" between input and output data through learning, and adaptively memorize the learning content in the weights of the network. That is, BP neural network has a high degree of self-learning and adaptive capabilities.

[0005] However, the BP neural network model also has some shortcomings. The BP network has local minimum values, and the selection of the initial weights and thresholds of the BP neural network is often randomly assigned. The BP neural network mainly uses gradient information to determine the search direction. The unreasonable initial weights and thresholds may cause the network to fall into the local optimum, thus causing the network training to fail.

[0006] Secondly, the quality and efficiency of case trials involve many indicators, many of which have strong linear correlations, so there will be some duplication in the information provided by the samples on these two dimensions. In addition, in order to reduce the amount of data calculation and remove noise, it is also required to remove the less important dimensions in the data, and the dimensions are unrelated.

[0007] In view of the drawbacks of data redundancy and BP neural network modeling, it is necessary to design a method to solve the above two problems. Summary of the invention

[0008] Purpose of the invention: The purpose of the present invention is to provide a case quality and efficiency evaluation method based on PCA-BP-GA, which introduces principal component analysis and genetic algorithm into the BP neural network to enable the model to obtain better results.

[0009] Another object of the present invention is to provide a case quality and efficiency evaluation system based on PCA-BP-GA.

[0010] Technical solution: A case quality and efficiency evaluation method based on PCA-BP-GA of the present invention comprises the following steps:

[0011] S1. Receive case information to be evaluated and organize it into case-related data sets, which include indicator data of case quality and efficiency systems of various courts;

[0012] S2. Preprocess the data set; use principal component analysis (PCA) to reduce the dimension of the data set and extract the main indicator data that affect the quality and efficiency of the case;

[0013] S3, inputting the main indicator data affecting case quality and efficiency extracted in step S2 into the case quality and efficiency evaluation model, and obtaining the quality and efficiency evaluation score of the corresponding case through the case quality and efficiency evaluation model;

[0014] S4. Complete the case evaluation work based on the case quality and efficiency evaluation scores.

[0015] Furthermore, the method for preprocessing the data set in step S2 is:

[0016] S201, data set X = {X 1 ,X 2 ,X 3 ,…,X r}, where r represents the number of courts in the data set, i.e. the number of samples in the data set, and X j represents the index data of the quality and efficiency system of the j-th court case, j = 1, 2, …, r, Among them, n represents the number of indicators of case quality and efficiency system of each court, represents the quality and efficiency system indicator score of the i-th case in the j-th court, i = 1, 2, …, n;

[0017] S202. Centralize the scoring of the quality and efficiency system indicators of all cases in each court: represents the average score of the case quality and efficiency system indicator of the jth sample;

[0018] S203, establish a covariance matrix: Among them, X T is the transposed matrix of the data set X;

[0019] S204, after obtaining the covariance matrix A, calculating the eigenvalues ​​and eigenvectors of the covariance matrix;

[0020] S205, arranging the eigenvectors into a matrix by row from top to bottom according to the corresponding eigenvalues, taking the first q rows to form a matrix P; P represents the quality and efficiency indicator data of each court case after dimensionality reduction, and q represents the number of quality and efficiency indicators of each court case after dimensionality reduction;

[0021] S206. Determine the final selected main indicator data P that affects the quality and efficiency of the case.

[0022] Furthermore, the case quality and efficiency evaluation model in step S3 is a trained improved BP neural network model, and the training steps of the improved BP neural network model are:

[0023] The initial BP neural network model is trained through training sample data, which are the main indicator data affecting case quality and efficiency after PCA dimensionality reduction;

[0024] According to the global error E of the initial BP neural network model, the genetic algorithm is used to update the connection weights and thresholds of the initial BP neural network model, and the training is completed to obtain the improved BP neural network model.

[0025] Furthermore, the training steps of the initial BP neural network model are:

[0026] S301, BP neural network initialization: the number of nodes in the input layer is n, the number of nodes in the hidden layer is l, and the number of nodes in the output layer is m; the weight from the input layer to the hidden layer is w ij , the weight from the hidden layer to the output layer is w jk , the threshold from the input layer to the hidden layer is a j , the threshold from the hidden layer to the output layer is b k ; The learning rate is λ, and the activation function is: The main indicator data P that affects the quality and efficiency of cases after PCA dimension reduction, the weight w from the input layer to the hidden layer ij And the threshold a from the input layer to the hidden layerj As the input of BP neural network;

[0027] S302, calculate the output of each hidden layer, the hidden layer output H j The calculation formula is: Where, i = 1, 2, ..., n, j = 1, 2, ..., l; It represents the quality and efficiency system index score of the i-th case of the j-th court, which serves as the input data of each node in the hidden layer;

[0028] S303, calculate the output layer output, the output layer output O k The calculation formula is: Where k = 1, 2, ..., m;

[0029] S304, using the least squares method to calculate the BP neural network error E, the calculation formula is: where Y k Refers to the expected output;

[0030] S305, if the BP neural network error E is greater than the set accuracy value, then update the weights and thresholds from the input layer to the hidden layer and the thresholds from the hidden layer to the output layer, and use the updated weights and thresholds to repeat the back propagation training until the BP neural network error E meets the accuracy ε, and obtain the trained initial BP neural network model;

[0031] S306: Normalize the output result, that is, set the output result range to (0, 1], as the preliminary quality and efficiency evaluation score.

[0032] Furthermore, the weight update formula from the input layer to the hidden layer in step S305 is: e k Represents the current error value, and the threshold update formula from the input layer to the hidden layer is: The threshold update formula from the hidden layer to the output layer is:

[0033] Furthermore, the training steps of the improved BP neural network model are:

[0034] S311, determine the initial population: set the initial population size, number of iterations, crossover probability, mutation probability of the genetic algorithm, and set the number of hidden layers, number of iterations, and learning rate of the BP neural network;

[0035] S312, running the initial BP neural network model to obtain a global error value E;

[0036] S313, calculate the fitness f(i), the calculation formula is

[0037] S314, select chromosomes with high fitness for replication, crossover, mutation, and termination;

[0038] S315, repeat S314 cyclically until the global error E meets the accuracy requirement or the number of iterations reaches a maximum;

[0039] S316, decode the optimal solution and assign it to the weight w from the input layer to the hidden layer ij , the bias from the input layer to the hidden layer is a j , start executing the initial BP neural network model, train the improved BP neural network model, and finally normalize the model output results, that is, the output result range is (0,1], as the final quality and efficiency evaluation score.

[0040] In another embodiment of the present invention, a case quality and efficiency evaluation system based on PCA-BP-GA includes:

[0041] The data receiving module is used to receive the case information to be evaluated and organize it into case-related data sets;

[0042] Data preprocessing module, used to reduce the dimension of case data sets and extract the main indicator data that affect case quality and efficiency;

[0043] The case quality and efficiency evaluation module inputs the main indicator data affecting the case quality and efficiency of the case to be evaluated after PCA dimensionality reduction into the case quality and efficiency evaluation model, obtains the quality and efficiency evaluation score of the case to be evaluated, and completes the case quality and efficiency evaluation.

[0044] Optionally, the data preprocessing module uses principal component analysis PCA to reduce the dimension of the data set; the case quality and efficiency evaluation module includes: a BP neural network initialization and training module, which is used to initialize the BP neural network parameters and perform preliminary training to obtain an initial BP neural network model; a BP neural network improvement module, which uses a genetic algorithm to update the connection weights and thresholds of the initial BP neural network model according to the global error E of the initial BP neural network model, and completes the training to obtain an improved BP neural network model.

[0045] In another embodiment of the present invention, an electronic device includes a memory and a processor, wherein:

[0046] A memory for storing computer programs that can be run on the processor;

[0047] The processor is used to execute the steps of the above-mentioned case quality and efficiency evaluation method based on PCA-BP-GA when running the computer program.

[0048] In another embodiment of the present invention, a storage medium stores a computer program, and when the computer program is executed by at least one processor, the steps of the case quality and efficiency evaluation method based on PCA-BP-GA are implemented as described above.

[0049] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0050] (1) The BP neural network is used to evaluate the quality and efficiency of cases, replacing the previous manual quality and efficiency evaluation method and enhancing the objectivity of the evaluation work. Secondly, the use of neural networks to evaluate the quality and efficiency of cases can continuously adjust the model according to the data in order to achieve the best adaptability.

[0051] (2) In order to further improve the speed of model building, the principal hierarchy analysis method (PCA) in machine learning is used to pre-process the data. The quality and efficiency of case trials involve many indicators and characteristics, and many indicators have strong linear correlations. In addition, in order to reduce the amount of data calculation and remove noise, it is also required to eliminate the less important dimensions in the data. Therefore, the use of PCA can solve the above-mentioned problems.

[0052] (3) Aiming at the local optimal problem of BP neural network, the initial weights and thresholds selected by genetic algorithm can reduce the redundancy of BP neural network and accelerate the convergence speed, so the use of genetic algorithm (GA) is reasonable and effective. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a flow chart of the method of the present invention;

[0054] Figure 2 It is a flow chart of initial BP neural network model training in the method of the present invention;

[0055] Figure 3 It is a flow chart of improving the BP neural network model training in the method of the present invention. DETAILED DESCRIPTION

[0056] The present invention is further described below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only a 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 creative work are within the scope of protection of the present invention.

[0057] Embodiment 1:

[0058] like Figure 1 As shown, the case quality and efficiency evaluation method based on PCA-BP-GA described in the present invention includes the following steps:

[0059] S1. Receive case information to be evaluated, organize it into case-related data sets, and determine case quality and efficiency system indicators;

[0060] In this embodiment, the case data information includes the work completion index data of each court, which is organized into a case-related data set, and the data set includes the case quality and efficiency system index data of each court, and the case quality and efficiency system index data includes the case quality and efficiency system index and the index score;

[0061] S2. Preprocess the data set X; use principal component analysis (PCA) to reduce the dimension of the data set and extract the main indicator data that affect the quality and efficiency of the case, that is, the main indicator characteristics and indicator scores that affect the quality and efficiency of the case. Examples of the main indicator data are as follows:

[0062] Table 1 Main indicators of the case

[0063]

[0064] Data preprocessing includes data cleaning and normalization. For data cleaning, since many data features affect the training efficiency, we decided to use PCA principal hierarchy analysis to reduce the data dimension. Specifically:

[0065] The method for preprocessing the dataset is:

[0066] S201, data set X = {X 1 ,X 2 ,X 3 ,…,X r}, where r represents the number of courts in the data set, i.e. the number of samples in the data set, and X j represents the index data of the quality and efficiency system of the j-th court case, j = 1, 2, …, r, Among them, n represents the number of indicators of case quality and efficiency system of each court, represents the quality and efficiency system indicator score of the i-th case in the j-th court, i = 1, 2, …, n;

[0067] S202. Centralize the scoring of the quality and efficiency system indicators of all cases in each court: represents the average score of the case quality and efficiency system indicator of the jth sample;

[0068] S203, establish a covariance matrix:

[0069]

[0070] in, is the transposed matrix of the data set X;

[0071] S204. After obtaining the covariance matrix A, calculate the eigenvalues ​​and eigenvectors of the covariance matrix, A = QΛQ A1 , where Λ is a diagonal matrix, each diagonal element is an eigenvalue, and Q is a matrix composed of the eigenvectors of the covariance matrix A;

[0072] Eigenvalues ​​of the covariance matrix: λ1≥λ2≥…≥λ p ≥0;

[0073] The eigenvector is: a1 = [a 11 ,a 21 …a p1 ],a2=[a 12 ,a 22 …a p2 ],…a p =[a 1p ,a 2p …a pp ];

[0074] S205, arranging the eigenvectors into a matrix by row from top to bottom according to the corresponding eigenvalues, taking the first q rows to form a matrix P; P represents the quality and efficiency indicator data of each court case after dimensionality reduction, and q represents the number of quality and efficiency indicators of each court case after dimensionality reduction;

[0075] S206, determining the main indicator data P that affects the quality and efficiency of the case that is finally selected as the input data of the neural network;

[0076] S3, inputting the main indicator data affecting case quality and efficiency extracted in step S2 into the case quality and efficiency evaluation model, and obtaining the quality and efficiency evaluation score of the corresponding case through the case quality and efficiency evaluation model;

[0077] The case quality and efficiency evaluation model is a trained improved BP neural network model. The training steps of the improved BP neural network model are:

[0078] The initial BP neural network model is trained through training sample data; specifically: the main indicator data P that affects the quality and efficiency of cases after PCA dimension reduction is imported, and the training samples and test samples are divided, and the weights w from the input layer to the hidden layer are randomly initialized. ij , the bias from the input layer to the hidden layer is a j , the weight w from the input layer to the hidden layer ij and the bias a from the input layer to the hidden layer j is the weight and threshold of the main indicator data P, and the main indicator data P that affects the quality and efficiency of the case after PCA dimensionality reduction, the weight w from the input layer to the hidden layer ij And the bias from the input layer to the hidden layer is a jAs the input of the BP neural network, calculate the output Y1 of each hidden layer of the BP neural network, and calculate the output Y2 of the output layer; thus, the output result Y of the entire BP neural network is obtained, and the error E of the BP neural network is calculated; adjust the weight w from the input layer to the hidden layer through learning ij And the bias from the input layer to the hidden layer is a j , repeat the training until the BP neural network error E reaches the required accuracy, and the initial BP neural network model is obtained;

[0079] According to the global error E of the initial BP neural network model, the genetic algorithm is used to adjust the weights w from the input layer to the hidden layer of the initial BP neural network model. ij And the bias from the input layer to the hidden layer is a j Update and complete the training to obtain the improved BP neural network model; Specifically: determine the initial population popu and encode it, run the initial BP neural network model trained in step S2, obtain the global error E to calculate the fitness, select chromosomes with high fitness to replicate, cross, mutate, generate a new population, until the accuracy condition is met, decode and assign the weight w from the new input layer to the hidden layer ij And the bias from the input layer to the hidden layer is a j The BP neural network is trained to obtain an improved BP neural network model.

[0080] The specific steps are:

[0081] S301, BP neural network initialization, the number of nodes in the input layer is n, the number of nodes in the hidden layer is l, the number of nodes in the output layer is m; the weight from the input layer to the hidden layer is w ij , the weight from the hidden layer to the output layer is w jk , the threshold from the input layer to the hidden layer is a j , the threshold from the hidden layer to the output layer is b k ; The learning rate is η, and the activation function is: The main indicator data P that affects the quality and efficiency of cases after PCA dimension reduction, the weight w from the input layer to the hidden layer ij And the threshold a from the input layer to the hidden layer j As the input of BP neural network;

[0082] w ij is the connection weight matrix, which is an n*n matrix with internal values ​​generated randomly. j It is a column vector of size n*1, and the internal values ​​are randomly generated.

[0083] S302, calculate the output of each hidden layer, the hidden layer output H j The calculation formula is: Where, i = 1, 2, ..., n, j = 1, 2, ..., l; It represents the quality and efficiency system index score of the i-th case of the j-th court, which serves as the input data of each node in the hidden layer;

[0084] S303, calculate the output layer output, the output layer output O k The calculation formula is: Where k = 1, 2, ..., m;

[0085] S304, using the least squares method to calculate the BP neural network error E, the calculation formula is: where Y k Refers to the expected output;

[0086] S305. If the BP neural network error E is greater than the set accuracy value, the weight and threshold from the input layer to the hidden layer and the threshold from the hidden layer to the output layer are updated; the weight update formula from the input layer to the hidden layer is: e k Represents the current error value, and the threshold update formula from the input layer to the hidden layer is: The threshold update formula from the hidden layer to the output layer is: The updated connection weights are used for repeated back propagation training until the BP neural network error E meets the accuracy ε, and the trained initial BP neural network model is obtained.

[0087] In this embodiment, the number of repeated training is set to 1000, the precision ε is set to 0.0001-0.01, and the learning rate is set to 0.01-0.1; until the BP neural network error E meets the precision ε, the trained initial BP neural network model is obtained;

[0088] S306. Normalize the output result, that is, set the output result range to (0,1], as the score for the preliminary case quality and efficiency evaluation.

[0089] Training of the improved BP neural network model: GA improves the BP neural network, determines the initial population, encodes it, runs the BP neural network, and obtains the global error E(i) to calculate the fitness. Select chromosomes with high fitness to replicate, crossover, and mutate, generate new populations, meet the accuracy conditions, decode and assign values, and use new W, b to train the BP neural network; the specific steps are:

[0090] S311, determine the initial population: set the genetic algorithm parameters as follows: the initial population size is 10, the number of iterations is 10, the crossover probability is 0.3, and the mutation probability is 0.1; the number of hidden layers of the BP neural network is set to 5, the number of iterations is 100, and the learning rate is 0.1;

[0091] S312, running the initial BP neural network model to obtain a global error value E;

[0092] S313, calculate the fitness f(i), the calculation formula is

[0093] S314, select chromosomes with high fitness for replication, crossover, mutation, and termination;

[0094] S315, repeat S314 cyclically until the global error E meets the accuracy requirement or the number of iterations reaches a maximum;

[0095] S316, decode the optimal solution and assign it to the weight w from the input layer to the hidden layer ij And the bias from the input layer to the hidden layer is a j , start executing the initial BP neural network model, train the improved BP neural network model, and finally normalize the model output results, that is, the output result range is (0,1], as the final case quality and efficiency evaluation score.

[0096] S4. Complete the case evaluation work based on the case quality and efficiency evaluation scores.

[0097] Embodiment 2:

[0098] In this embodiment, a case quality and efficiency evaluation system based on PCA-BP-GA includes:

[0099] The data receiving module is used to receive the case information to be evaluated and organize it into case-related data sets;

[0100] The data preprocessing module is used to reduce the dimension of case-related data sets and extract the main indicator data that affect the quality and efficiency of the case; in this embodiment, the data preprocessing module uses principal component analysis PCA to reduce the dimension of the data set;

[0101] The case quality and efficiency evaluation module inputs the main indicator data affecting the case quality and efficiency of the case to be evaluated after PCA dimensionality reduction into the case quality and efficiency evaluation model, obtains the quality and efficiency evaluation score of the case to be evaluated, and completes the case quality and efficiency evaluation.

[0102] The case quality and efficiency evaluation module includes: a BP neural network initialization and training module, which is used to initialize the BP neural network parameters and perform preliminary training to obtain the initial BP neural network model; a BP neural network improvement module, which uses a genetic algorithm to update the connection weights and thresholds of the initial BP neural network model according to the global error E of the initial BP neural network model, and completes the training to obtain the improved BP neural network model.

[0103] Embodiment 3:

[0104] In this embodiment, a device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, including a memory and a processor, wherein:

[0105] A memory for storing computer programs that can be run on the processor;

[0106] The processor is used to execute the steps of the above-mentioned case quality and efficiency evaluation method based on PCA-BP-GA when running the computer program, and achieve the technical effect consistent with the above-mentioned method.

[0107] Embodiment 4:

[0108] This embodiment also provides a storage medium, on which a computer program is stored. When the computer program is executed by at least one processor, the steps of the above-mentioned case quality and efficiency evaluation method based on PCA-BP-GA are implemented, and the technical effect consistent with the above-mentioned method is achieved.

[0109] In summary, the present invention proposes to use BP neural network modeling to complete the construction of the quality and efficiency evaluation model, and then realize the quality and efficiency evaluation work. In the whole process, the samples of the data set are input into the network for training. After the accuracy set by the network is reached, the purpose of modeling is achieved. After the modeling is completed, the new samples are input into the network, and the quality and efficiency evaluation of the samples can be realized.

Claims

1. A case quality and efficiency evaluation method based on PCA-BP-GA, characterized in that: The following steps are involved: S1. Receive case information to be evaluated and organize it into case-related data sets, which include indicator data of case quality and efficiency systems of various courts; S2. Preprocess the data set; use principal component analysis (PCA) to reduce the dimension of the data set and extract the main indicator data that affect the quality and efficiency of the case; S3. Input the main indicator data affecting case quality and efficiency extracted in step S2 into the case quality and efficiency evaluation model, and obtain the quality and efficiency evaluation score of the corresponding case through the case quality and efficiency evaluation model; the case quality and efficiency evaluation model is a trained improved BP neural network model, and the training steps of the improved BP neural network model are: The initial BP neural network model is trained through training sample data, which are the main indicator data affecting case quality and efficiency after PCA dimensionality reduction; According to the global error E of the initial BP neural network model, the genetic algorithm is used to update the connection weights and thresholds of the initial BP neural network model, and the training is completed to obtain the improved BP neural network model; The training steps of the initial BP neural network model are: S301, BP neural network initialization: the number of nodes in the input layer is n, the number of nodes in the hidden layer is l, and the number of nodes in the output layer is m; the weight from the input layer to the hidden layer is w ij , the weight from the hidden layer to the output layer is w jk , the threshold from the input layer to the hidden layer is a j , the threshold from the hidden layer to the output layer is b k ; The learning rate is λ, and the activation function is: The main indicator data P that affects the quality and efficiency of cases after PCA dimension reduction, the weight w from the input layer to the hidden layer ij And the threshold a from the input layer to the hidden layer j As the input of BP neural network; S302, calculate the output of each hidden layer, the hidden layer output H j The calculation formula is: Where, i = 1, 2, ..., n, j = 1, 2, ..., l; It represents the quality and efficiency system index score of the i-th case in the j-th court, which serves as the input data of each node in the hidden layer; S303, calculate the output layer output, the output layer output O k The calculation formula is: Where k = 1, 2, ..., m; S304, using the least squares method to calculate the BP neural network error E, the calculation formula is: where Y k Refers to the expected output; S305, if the BP neural network error E is greater than the set accuracy value, then update the weights and thresholds from the input layer to the hidden layer and the thresholds from the hidden layer to the output layer, and use the updated weights and thresholds to repeat the back propagation training until the BP neural network error E meets the accuracy ε, and obtain the trained initial BP neural network model; The weight update formula from the input layer to the hidden layer is: e k Represents the current error value, and the threshold update formula from the input layer to the hidden layer is: The threshold update formula from the hidden layer to the output layer is: S306, normalizing the output result, that is, the output result range is (0,1], as the preliminary quality and efficiency evaluation score; The training steps of the improved BP neural network model are: S311, determine the initial population: set the initial population size, number of iterations, crossover probability, mutation probability of the genetic algorithm, and set the number of hidden layers, number of iterations, and learning rate of the BP neural network; S312, running the initial BP neural network model to obtain a global error value E; S313, calculate the fitness f(i), the calculation formula is S314, select chromosomes with high fitness for replication, crossover, mutation, and termination; S315, repeat S314 cyclically until the global error E meets the accuracy requirement or the number of iterations reaches a maximum; S316, decode the optimal solution and assign it to the weight w from the input layer to the hidden layer ij , the threshold a from the input layer to the hidden layer j , start executing the initial BP neural network model, train the improved BP neural network model, and finally normalize the model output results, that is, the output result range is (0,1], as the final quality and efficiency evaluation score; S4. Complete the case evaluation work based on the case quality and efficiency evaluation scores.

2. According to claim 1, a case quality and efficiency evaluation method based on PCA-BP-GA is characterized in that: The method for preprocessing the data set in step S2 is: S201, data set X = {X 1 ,X 2 ,X 4 ,…,X 6 }, where r represents the number of courts in the data set, i.e. the number of samples in the data set, and X j represents the index data of the quality and efficiency system of the j-th court case, j = 1, 2, …, r, Among them, n represents the number of indicators of case quality and efficiency system of each court, represents the quality and efficiency system indicator score of the i-th case in the j-th court, i = 1, 2, …, n; S202. Centralize the scoring of the quality and efficiency system indicators of all cases in each court: represents the average score of the case quality and efficiency system indicator of the jth sample; S203, establish a covariance matrix: in, is the transposed matrix of the data set X; S204, after obtaining the covariance matrix A, calculating the eigenvalues ​​and eigenvectors of the covariance matrix; S205, arranging the eigenvectors into a matrix by row from top to bottom according to the corresponding eigenvalues, taking the first q rows to form a matrix P; P represents the quality and efficiency indicator data of each court case after dimensionality reduction, and q represents the number of quality and efficiency indicators of each court case after dimensionality reduction; S206. Determine the final selected main indicator data P that affects the quality and efficiency of the case.

3. A case quality and efficiency evaluation system based on PCA-BP-GA, characterized by: include: The data receiving module is used to receive the case information to be evaluated and organize it into case-related data sets; Data preprocessing module, used to reduce the dimension of case data sets and extract the main indicator data that affect case quality and efficiency; The case quality and efficiency evaluation module inputs the main indicator data affecting the case quality and efficiency after PCA dimension reduction into the case quality and efficiency evaluation model, obtains the quality and efficiency evaluation score of the case to be evaluated, and completes the case quality and efficiency evaluation: The case quality and efficiency evaluation model is a trained improved BP neural network model. The training steps of the improved BP neural network model are: The initial BP neural network model is trained through training sample data, which are the main indicator data affecting case quality and efficiency after PCA dimensionality reduction; According to the global error E of the initial BP neural network model, the genetic algorithm is used to update the connection weights and thresholds of the initial BP neural network model, and the training is completed to obtain the improved BP neural network model; The training steps of the initial BP neural network model are: S301, BP neural network initialization: the number of nodes in the input layer is n, the number of nodes in the hidden layer is l, and the number of nodes in the output layer is m; the weight from the input layer to the hidden layer is w ij , the weight from the hidden layer to the output layer is w jk , the threshold from the input layer to the hidden layer is a j , the threshold from the hidden layer to the output layer is b k ; The learning rate is λ, and the activation function is: The main indicator data P that affects the quality and efficiency of cases after PCA dimension reduction, the weight w from the input layer to the hidden layer ij And the threshold a from the input layer to the hidden layer j As the input of BP neural network; S302, calculate the output of each hidden layer, the hidden layer output H j The calculation formula is: Where, i = 1, 2, ..., n, j = 1, 2, ..., l; It represents the quality and efficiency system index score of the i-th case in the j-th court, which serves as the input data of each node in the hidden layer; S303, calculate the output layer output, the output layer output O k The calculation formula is: Where k = 1, 2, ..., m; S304, using the least squares method to calculate the BP neural network error E, the calculation formula is: where Y k Refers to the expected output; S305, if the BP neural network error E is greater than the set accuracy value, then update the weights and thresholds from the input layer to the hidden layer and the thresholds from the hidden layer to the output layer, and use the updated weights and thresholds to repeat the back propagation training until the BP neural network error E meets the accuracy ε, and obtain the trained initial BP neural network model; The weight update formula from the input layer to the hidden layer is: e k Represents the current error value, and the threshold update formula from the input layer to the hidden layer is: The threshold update formula from the hidden layer to the output layer is: S306, normalizing the output result, that is, the output result range is (0,1], as the preliminary quality and efficiency evaluation score; The training steps of the improved BP neural network model are: S311, determine the initial population: set the initial population size, number of iterations, crossover probability, mutation probability of the genetic algorithm, and set the number of hidden layers, number of iterations, and learning rate of the BP neural network; S312, running the initial BP neural network model to obtain a global error value E; S313, calculate the fitness f(i), the calculation formula is S314, select chromosomes with high fitness for replication, crossover, mutation, and termination; S315, repeat S314 cyclically until the global error E meets the accuracy requirement or the number of iterations reaches a maximum; S316, decode the optimal solution and assign it to the weight w from the input layer to the hidden layer ij , the threshold a from the input layer to the hidden layer j , start executing the initial BP neural network model, train the improved BP neural network model, and finally normalize the model output results, that is, the output result range is (0,1], as the final quality and efficiency evaluation score.

4. According to claim 3, a case quality and efficiency evaluation system based on PCA-BP-GA is characterized in that: The data preprocessing module uses principal component analysis (PCA) to reduce the dimension of the data set; The case quality and efficiency evaluation module includes: a BP neural network initialization and training module, which is used to initialize the BP neural network parameters and perform preliminary training to obtain an initial BP neural network model; a BP neural network improvement module, which uses a genetic algorithm to update the connection weights and thresholds of the initial BP neural network model based on the global error E of the initial BP neural network model, and completes the training to obtain an improved BP neural network model.

5. An electronic device, characterized in that: comprising a memory and a processor, wherein: A memory for storing computer programs that can be executed on the processor; A processor is used to execute the steps of a case quality and efficiency evaluation method based on PCA-BP-GA as described in any one of claims 1-2 when running the computer program.

6. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by at least one processor, implements the steps of a case quality and efficiency evaluation method based on PCA-BP-GA as described in any one of claims 1-2.

Images