Rural domestic sewage treatment mode optimization method based on neural network

Abstract

The invention discloses a rural domestic sewage treatment mode optimization method based on a neural network, and relates to the technical field of sewage treatment, an optimization system collects historical data related to rural domestic sewage treatment, and the feature data having significant influence on the sewage treatment effect in the historical data is screened out through a principal component analysis method; the neural network model analyzes the feature data and then outputs a plurality of treatment modes, after the plurality of treatment modes are subjected to loop iteration through a genetic algorithm, a plurality of optimized treatment schemes are generated, the generated treatment schemes are sorted to obtain a scheme list, and the first treatment scheme in the scheme list is selected as the treatment scheme of the rural domestic sewage. According to the optimization system, the complex relation among all factors in the sewage treatment process is captured through the neural network, then the optimal treatment mode adapting to the current sewage treatment condition is automatically generated, a more accurate optimization result is provided, and the response speed and the stability of the treatment effect are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of sewage treatment, and in particular to a method for optimizing rural domestic sewage treatment modes based on a neural network. Background Art

[0002] Rural domestic sewage treatment is an important measure to improve the rural living environment, protect farmers' health and protect water resources. The discharge of rural domestic sewage has increased year by year. However, due to the weak infrastructure and imperfect management system in rural areas, the sewage treatment level in many places is low, or even there is no effective sewage treatment system, which leads to water pollution, soil degradation and ecological environment deterioration. The prior art has the following defects: 1. Traditional sewage treatment optimization systems often rely on linear models or simple empirical formulas, which cannot effectively deal with the complexity and nonlinear problems involved in the sewage treatment process. For example, the complex impact of different geographical conditions, climate factors and sewage characteristics on treatment effects is difficult to accurately model through traditional methods; 2. As it is difficult to accurately model traditional sewage treatment optimization systems through traditional methods, traditional sewage treatment optimization systems usually rely on manual experience or static rules for decision-making, lacking flexibility and real-time response capabilities. This approach may not be able to respond to emergencies or environmental changes in a timely manner, resulting in unstable treatment effects.

[0003] Based on this, the present invention proposes a rural domestic sewage treatment model optimization method based on neural network. The neural network captures the complex relationship between various factors in the sewage treatment process, and then automatically generates the optimal treatment model that adapts to the current sewage treatment conditions. It not only provides more accurate optimization results, but also improves the response speed and the stability of the treatment effect. Summary of the invention

[0004] The purpose of the present invention is to provide a rural domestic sewage treatment mode optimization method based on neural network to solve the shortcomings of the background technology.

[0005] In order to achieve the above-mentioned purpose, the present invention provides the following technical solution: a rural domestic sewage treatment model optimization method based on neural network.

[0006] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. The present invention collects historical data related to rural domestic sewage treatment through an optimization system, and screens out characteristic data that has a significant impact on sewage treatment effects in the historical data through a principal component analysis method. After analyzing the characteristic data, the neural network model outputs several treatment modes, and after iterating the several treatment modes through a genetic algorithm, generates several optimized treatment plans, sorts the generated treatment plans to obtain a plan list, and selects the treatment plan ranked first in the plan list as the treatment plan for rural domestic sewage. The optimization system captures the complex relationship between various factors in the sewage treatment process through a neural network, and then automatically generates the optimal treatment mode that adapts to the current sewage treatment situation, which not only provides more accurate optimization results, but also improves the response speed and stability of the treatment effect.

[0007] 2. In the actual treatment process, the present invention regularly evaluates whether the treatment effect of the current treatment plan meets the standard. If it does not meet the standard, other treatment plans are reselected based on the plan list to further ensure the treatment effect of rural domestic sewage. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0009] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0010] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 part of the embodiments of the present invention, not all of 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.

[0011] Example 1: Please refer to Figure 1 As shown, the rural domestic sewage treatment mode optimization method based on neural network described in this embodiment includes the following steps: The optimization system collects historical data related to rural domestic sewage treatment, including sewage discharge, water quality parameters (such as COD, BOD, ammonia nitrogen, total phosphorus), treatment facility operation data, geographical environment data, population density, etc. The principal component analysis method is used to screen out characteristic data that have a significant impact on the sewage treatment effect in the historical data, such as terrain conditions, climate factors, residents' water use habits, etc. The characteristic data are substituted into the pre-trained neural network model. The neural network model analyzes the characteristic data and outputs several treatment modes. After iterating several treatment modes through a genetic algorithm, several optimized treatment plans are generated. The generated treatment plans are sorted to obtain a list of plans, and the treatment plan ranked first in the list of plans is selected as the treatment plan for rural domestic sewage. In the actual treatment process, the treatment effect of the current treatment plan is regularly evaluated to see if it meets the standards. If it does not meet the standards, other treatment plans are reselected based on the list of plans.

[0012] This application collects historical data related to rural domestic sewage treatment through an optimization system, and screens out characteristic data that has a significant impact on sewage treatment effects in historical data through principal component analysis. After analyzing the characteristic data, the neural network model outputs several treatment modes, and after iterating the several treatment modes through a genetic algorithm, generates several optimized treatment plans, sorts the generated treatment plans to obtain a list of plans, and selects the treatment plan ranked first in the list of plans as the treatment plan for rural domestic sewage. The optimization system captures the complex relationship between various factors in the sewage treatment process through a neural network, and then automatically generates the optimal treatment mode that adapts to the current sewage treatment situation, which not only provides more accurate optimization results, but also improves the response speed and stability of the treatment effect.

[0013] During the actual treatment process, this application will regularly evaluate whether the treatment effect of the current treatment plan meets the standards. If it does not meet the standards, other treatment plans will be reselected based on the plan list to further ensure the treatment effect of rural domestic sewage.

[0014] The optimization system in this application includes a data collection and processing module, a governance solution output module, and a solution evaluation module; Data collection and processing module: collect historical data related to rural domestic sewage treatment, including sewage discharge, water quality parameters (such as COD, BOD, ammonia nitrogen, total phosphorus), treatment facility operation data, geographical environment data, population density, etc., and select characteristic data that have a significant impact on sewage treatment effects from historical data through principal component analysis and other methods, such as terrain conditions, climate factors, residents' water use habits, etc., and send the characteristic data to the treatment plan output module; Governance plan output module: Substitute the feature data into the pre-trained neural network model. The neural network model analyzes the feature data and outputs several governance modes. After iterating the several governance modes through the genetic algorithm, several optimized governance plans are generated. The generated governance plans are sorted to obtain a plan list, which is sent to the plan evaluation module. Program evaluation module: Select the first treatment program in the program list as the treatment program for rural domestic sewage. In the actual treatment process, regularly evaluate whether the treatment effect of the current treatment program meets the standards. If not, reselect other treatment programs based on the program list.

[0015] Example 2: The optimization system collects historical data related to rural domestic sewage treatment, including sewage discharge, water quality parameters (such as COD, BOD, ammonia nitrogen, total phosphorus), treatment facility operation data, geographical environment data, population density, etc., including the following steps: Collect sewage discharge data: obtain sewage discharge information at different time periods and locations, and record daily or hourly discharge changes.

[0016] Collect water quality parameter data: Obtain historical data on key pollutants such as COD, BOD, nitrogen, and phosphorus from water quality monitoring stations or laboratories, including regular test records.

[0017] Collect treatment facility operation data: Collect historical operation data of sewage treatment facilities, such as treatment volume, energy consumption, equipment operation status, fault records, maintenance records, etc.

[0018] Collect geographic environmental data: Use GIS or other geographic information platforms to obtain environmental data such as topography, soil type, climate conditions, and water distribution in rural areas.

[0019] Collect population density data: Obtain information on historical population density changes in the target area from the Bureau of Statistics or census data.

[0020] Data cleaning: Process missing values, outliers, and noise in the data. Use interpolation, regression analysis, or adjacent point completion to fill in missing values ​​and filter or correct abnormal data.

[0021] Data format conversion: Convert data from different sources into a unified format, such as a unified timestamp, coordinate system or unit standard, to ensure data comparability and compatibility.

[0022] Data normalization: Normalize various types of data and standardize data of different dimensions to make them suitable for subsequent model input.

[0023] Data merging and integration: Merge different types of data according to time, space and other dimensions to build a comprehensive historical database.

[0024] Database design: Design a suitable database structure to efficiently store and manage various historical data. You can choose a relational database, time series database or big data platform for storage.

[0025] Data indexing and querying: Create indexes for data and optimize query performance, so that the system can quickly access and extract required historical data.

[0026] Data security and backup: Implement data security measures, such as encrypted storage, access control, etc., and perform regular data backup to ensure the security and integrity of historical data.

[0027] Data integrity check: Evaluate the integrity of the data to ensure the continuity of various types of data in time series and coverage in spatial distribution.

[0028] Data accuracy verification: Evaluate the accuracy of the data through cross-validation or comparison with other authoritative data sources.

[0029] Data consistency assessment: Check whether there are any contradictions or inconsistencies between data from different sources to ensure the reliability of the data.

[0030] The principal component analysis method is used to screen out characteristic data in historical data that have a significant impact on the sewage treatment effect, such as terrain conditions, climate factors, residents' water use habits, etc., including the following steps: The principal component analysis method is sensitive to the scale of the data, so the data needs to be standardized, all feature data are converted to the same scale, the mean of each feature is subtracted, and divided by its standard deviation, so that the mean of the data distribution of each feature is 0 and the standard deviation is 1; Calculate the covariance matrix of the standardized data set. Each element of the covariance matrix represents the covariance between two features. Perform eigenvalue decomposition on the covariance matrix to obtain a set of eigenvalues ​​and corresponding eigenvectors. The eigenvalue represents the variance in the direction of the corresponding eigenvector, and the eigenvector represents the direction of the principal component. Sort the eigenvectors according to the size of the eigenvalues, select the first few principal components whose cumulative explained variance reaches 80%-90%, and determine the number of principal components to be selected by the cumulative variance contribution rate (that is, the ratio of the sum of the eigenvalues ​​of the first few principal components selected to the sum of all eigenvalues); Multiply the standardized data set by the eigenvector matrix of the selected principal component to obtain the reduced-dimensional data. The reduced-dimensional data set is equal to the standardized data set multiplied by the eigenvector matrix. Analyze the projection coefficients of each principal component in the original feature space. The features corresponding to larger coefficients have a greater impact on the principal component. Select the original features whose projection coefficients are greater than the coefficient threshold as the key features.

[0031] The technical solution of screening out characteristic data having a significant impact on sewage treatment effects from historical data by principal component analysis method belongs to the prior art and will not be elaborated in this application.

[0032] Substitute the feature data into the pre-trained neural network model. The neural network model analyzes the feature data and outputs several governance models, including the following steps: The feature data is input into the input layer of the neural network model in the form of batch processing or single input. The neural network model starts forward propagation and processes the input data layer by layer through the weights and activation functions of each layer. According to the feature data, the neural network model identifies different governance modes. The neural network model will give multiple governance modes at the output layer. The governance mode content includes the selection of governance equipment, the startup time of different governance equipment, and the parameter setting of the governance equipment.

[0033] If the amount of data is small or only one data instance needs to be processed, the instance is directly input into the model. For large-scale data sets, the data is input into the model in batches. Batch processing can speed up the calculation and optimize the model training process. The input layer is usually a simple linear layer that receives input data and prepares for subsequent calculations. The input data is directly passed to the first hidden layer of the network. Through the calculation of each hidden layer, the input data is subjected to feature extraction and complex pattern recognition. In the hidden layer, the neurons of each layer are fully connected with the neurons of the previous layer. The input data is subjected to matrix multiplication calculation with the corresponding weight matrix. A bias term is added to the output of each neuron to improve the expressive power of the model.

[0034] For Layer, the calculation formula is: , where is the weight matrix, is the bias term, The last output.

[0035] Introducing nonlinearity enables the model to learn and express complex nonlinear relationships. The linear combination of hidden layer calculations is transformed nonlinearly through the activation function to obtain the output of the layer. Layer, the activation function application formula is: , is the activation function.

[0036] Through forward propagation, data is processed and passed layer by layer to extract deeper features. Starting from the first hidden layer, the output of each layer is calculated and used as the input of the next layer until the final output layer. As the number of layers increases, the neural network gradually extracts high-level features of the input data and forms a deep understanding of the data.

[0037] Similar to the hidden layer, the output layer also linearly combines the output of the previous layer through the weight matrix. Common activation functions of the output layer include Softmax (for classification problems) and linear activation (for regression problems). The calculation formula of the output layer is: ,in For the last layer, is the output of the model.

[0038] After a number of treatment modes are iterated through a genetic algorithm, a number of optimized treatment plans are generated, the generated treatment plans are sorted to obtain a plan list, and the treatment plan ranked first in the plan list is selected as the rural domestic sewage treatment plan, including the following steps: Obtain all governance modes, calculate the fitness value of each governance mode according to the fitness function, select some governance modes according to the fitness value using the roulette selection method, perform randomization on the selected governance modes, repeat the fitness value calculation, roulette selection and randomization steps on the governance modes after the randomization operation, and output multiple groups of governance modes when the convergence conditions are met, and use the output governance modes as the governance plan.

[0039] All treatment plans are sorted from large to small according to the fitness value, a plan list is generated, and the treatment plan ranked first in the plan list is selected as the treatment plan for rural domestic sewage.

[0040] Calculating the fitness value of each governance model based on the fitness function includes the following steps: Obtain the pollutant treatment efficiency, operating cost and environmental impact index of the governance model; The pollutant treatment efficiency, operating cost and environmental impact index are normalized so that their value ranges are mapped to [0,1] to obtain the normalized value of pollutant treatment efficiency, the normalized value of operating cost and the normalized value of environmental impact index. The fitness value is obtained by subtracting the normalized value of operating cost from the normalized value of environmental impact index from the normalized value of pollutant treatment efficiency. The larger the fitness value, the better the overall performance of the governance model, and the more it should be selected.

[0041] The calculation expression of pollutant treatment efficiency is: , where is the pollutant treatment efficiency, The pollutant concentration of the sewage before treatment in the governance model, It is the pollutant concentration of sewage after treatment by the governance model. The greater the pollutant treatment efficiency, the better the treatment effect of the governance model on sewage.

[0042] The calculation expression of environmental impact index is: , where is the environmental impact index, For the control mode of pollutant gas emissions, The degree of secondary pollution of water bodies by the governance model. Solid waste emissions for governance models, is the noise level when the governance mode is running, , , , It is the weight coefficient of each influencing factor, and the weight coefficient is greater than 0. The larger the environmental impact index is, the greater the impact of the governance model on the environment when it is in operation.

[0043] According to the fitness value, some governance modes are selected using the roulette wheel selection method, and the selected governance modes are randomized. The fitness value calculation, roulette wheel selection and randomization steps are repeated for the governance modes after the randomization operation. When the convergence conditions are met, multiple groups of governance modes are output, and the output governance modes are used as the governance plan, including the following steps: The fitness values ​​of all governance modes are summed up to obtain the total fitness value. The fitness value of each governance mode is divided by the total fitness value to obtain the selection probability of the governance mode. The selection probability is mapped to the area corresponding to the virtual roulette wheel. The greater the selection probability of the governance mode, the larger the sector area occupied on the virtual roulette wheel. The virtual roulette wheel is started. When the virtual roulette wheel stops, the governance mode pointed to by the virtual pointer is selected. When the number of selected governance modes is equal to the quantity threshold, the selection is stopped. This step can improve the robustness of the optimization system. The selected governance model is randomized. The randomization operation includes crossover and mutation. The crossover operation generates a new governance model (child) by combining two different governance models (parents). This is to introduce new feature combinations to explore different solutions. The mutation operation generates new features by randomly modifying one or more genes in the governance model, increasing the diversity of the population and preventing it from falling into a local optimal solution.

[0044] When the number of iterations is equal to the number threshold, it is determined that the convergence condition is met.

[0045] In the actual governance process, regularly evaluate whether the governance effect of the current governance plan meets the standards. If not, reselect other governance plans based on the plan list, including the following steps: In the actual governance process, obtain real-time pollutant treatment efficiency, expected cost deviation and environmental impact index; Among them, the calculation logic of the expected cost deviation is: subtract the current expected cost of the governance model from the current actual operating cost of the governance model to obtain the expected cost deviation. The larger the expected cost deviation, the more the governance model deviates from the expected cost.

[0046] Substitute the pollutant treatment efficiency, expected cost deviation and environmental impact index into the formula to calculate the governance coefficient, which is expressed as: , where is the governance coefficient, is the environmental impact index, is the pollutant treatment efficiency, is the expected cost deviation, , , are the pollutant treatment efficiency, environmental impact index, and the proportional coefficient of expected cost deviation, respectively, and , , All are greater than 0; If the governance coefficient is greater than or equal to the standard threshold, the governance effect of the current governance plan is evaluated to be up to standard. If the governance coefficient is less than the standard threshold, the governance effect of the current governance plan is evaluated to be not up to standard. The second governance plan in the positive order in the plan list is selected for use. If the governance effect of the second governance plan is not up to standard, the governance plan is selected for use in the same way.

[0047] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0048] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0049] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. Optimization method of rural domestic sewage treatment mode based on neural network, Features: The optimization method comprises the following steps: The optimization system collects historical data related to rural domestic sewage treatment, and uses the principal component analysis method to screen out the characteristic data that affects the sewage treatment effect in the historical data. The characteristic data is substituted into the pre-trained neural network model, and the neural network model analyzes the characteristic data and outputs several treatment modes; After a number of treatment modes are iterated through a genetic algorithm, a number of optimized treatment plans are generated, the generated treatment plans are sorted to obtain a plan list, and the treatment plan ranked first in the plan list is selected as the treatment plan for rural domestic sewage; In the actual governance process, regularly evaluate whether the governance effect of the current governance plan meets the standards. If not, reselect other governance plans based on the plan list.

2. According to claim 1, the rural domestic sewage treatment mode optimization method based on neural network, Features: Substitute the feature data into the pre-trained neural network model. The neural network model analyzes the feature data and outputs several governance models, including the following steps: The feature data is input into the input layer of the neural network model in the form of batch processing or single input. The neural network model starts forward propagation and processes the input data layer by layer through the weights and activation functions of each layer. According to the feature data, the neural network model gives several governance modes at the output layer. The governance mode includes the selection of governance equipment, the startup time of different governance equipment, and the parameter setting of governance equipment.

3. According to claim 2, the rural domestic sewage treatment mode optimization method based on neural network, Features: After several governance models are iterated through genetic algorithms, several optimized governance solutions are generated, including the following steps: Obtain all governance modes, calculate the fitness value of each governance mode according to the fitness function, select some governance modes according to the fitness value using the roulette selection method, perform randomization on the selected governance modes, repeat the fitness value calculation, roulette selection and randomization steps on the governance modes after randomization, and output multiple groups of governance modes when the convergence conditions are met, and use the output governance modes as the governance plan; All treatment plans are sorted from large to small according to the fitness value, a plan list is generated, and the treatment plan ranked first in the plan list is selected as the treatment plan for rural domestic sewage.

4. According to claim 3, the rural domestic sewage treatment mode optimization method based on neural network, Features: Calculating the fitness value of each governance model based on the fitness function includes the following steps: The pollutant treatment efficiency, operating cost and environmental impact index of the governance model are obtained, and the pollutant treatment efficiency, operating cost and environmental impact index are normalized so that the value range of the pollutant treatment efficiency, operating cost and environmental impact index is mapped to [0,1], and the normalized value of the pollutant treatment efficiency, the normalized value of the operating cost and the normalized value of the environmental impact index are obtained. The fitness value is obtained by subtracting the normalized value of the operating cost from the normalized value of the environmental impact index from the normalized value of the pollutant treatment efficiency.

5. According to claim 4, the rural domestic sewage treatment mode optimization method based on neural network, Features: The roulette wheel selection method is used to select some governance modes according to the fitness value, and the selected governance modes are randomized, including the following steps: The fitness values ​​of all governance modes are summed up to obtain the total fitness value. The fitness value of each governance mode is divided by the total fitness value to obtain the selection probability of the governance mode. The selection probability is mapped to the area corresponding to the virtual roulette wheel. The greater the selection probability of the governance mode, the larger the sector area occupied on the virtual roulette wheel. The virtual roulette wheel is started. When the virtual roulette wheel stops, the governance mode pointed to by the virtual pointer is selected. The selection stops when the number of selected governance modes equals the quantity threshold. The selected governance model is subjected to randomization operations, which include crossover operations and mutation operations. The crossover operation generates a new governance model by combining two different governance models, and the mutation operation randomly modifies one or more contents in the governance model.

6. The rural domestic sewage treatment mode optimization method based on neural network according to claim 5, Features: In the actual governance process, regular evaluation is conducted to determine whether the governance effect of the current governance plan meets the standards, including the following steps: In the actual governance process, obtain real-time pollutant treatment efficiency, expected cost deviation and environmental impact index; Substitute the pollutant treatment efficiency, expected cost deviation and environmental impact index into the formula to calculate the governance coefficient, which is expressed as: , where is the governance coefficient, is the environmental impact index, is the pollutant treatment efficiency, is the expected cost deviation, , , are the pollutant treatment efficiency, environmental impact index, and the proportional coefficient of expected cost deviation, respectively, and , , All are greater than 0; If the governance coefficient is greater than or equal to the standard threshold, the governance effect of the current governance plan is evaluated to be up to standard. If the governance coefficient is less than the standard threshold, the governance effect of the current governance plan is evaluated to be not up to standard, and the second governance plan in the positive order in the plan list is selected for use.

7. The rural domestic sewage treatment mode optimization method based on neural network according to claim 4, Features: The calculation expression of the pollutant treatment efficiency is: , where is the pollutant treatment efficiency, The pollutant concentration of the sewage before treatment in the governance model, The pollutant concentration of the wastewater after treatment in the governance model; The calculation expression of the environmental impact index is: , where is the environmental impact index, For the control mode of pollutant gas emissions, The degree of secondary pollution of water bodies by the governance model. Solid waste emissions for governance models, is the noise level when the governance mode is running, , , , is the weight coefficient of each influencing factor.

8. The rural domestic sewage treatment mode optimization method based on neural network according to claim 7, Features: The principal component analysis method is used to screen out characteristic data that has a significant impact on the sewage treatment effect in historical data, including the following steps: Standardize the historical data, convert all feature data to the same scale, subtract the mean of each feature and divide it by its standard deviation, so that the mean of the data distribution of each feature is 0 and the standard deviation is 1; Calculate the covariance matrix of the standardized data set. Each element of the covariance matrix represents the covariance between two features. Perform eigenvalue decomposition on the covariance matrix to obtain a set of eigenvalues ​​and corresponding eigenvectors. The eigenvalue represents the variance in the direction of the corresponding eigenvector, and the eigenvector represents the direction of the principal component. Sort the eigenvectors according to the size of the eigenvalues, and select the first few principal components whose cumulative explained variance reaches 80%-90%; The standardized data set is multiplied by the eigenvector matrix of the selected principal component to obtain the reduced-dimensional data. The projection coefficients of each principal component in the original feature space are analyzed, and the original features whose projection coefficients are greater than the coefficient threshold are selected as key features.

Images