A background event identification method based on random forest classification model

Abstract

The present invention discloses a background event identification method based on a random forest classification model, comprising obtaining an original data set by measuring a high-activity standard sample and a blank sample through a detection system, preprocessing the original data set to obtain a standard data set; constructing a random forest classification model based on the standard data set, training the random forest classification model, and optimizing the random forest classification model based on the training error; using the optimized classification model to identify background events on a sample to be tested, performing a time series relationship analysis on the sample to be tested and the standard data set, and outputting an identification result. This method can not only improve the accuracy of background event identification, but also has good interpretability and can be directly applied to a background event identification system.

Description

Technical Field

[0001] The present invention relates to the field of event identification, and in particular to a background event identification method based on a random forest classification model. Background Art

[0002] In low-background liquid scintillation measurements, to achieve accurate measurements of even lower-activity samples, it is essential to further reduce the background level of the detection system while maintaining the highest possible detection efficiency. Traditional methods for reducing background include using low-background liquid scintillation detectors with enhanced passive shielding and employing time-coincidence and anti-coincidence techniques to reduce dark current noise from photomultiplier tubes and background events caused by cosmic rays or ambient gamma rays. In addition, signal processing-based pulse discrimination methods are widely used. For example, these methods utilize the pulse length-short gate integration ratio and the ratio of pulse amplitudes across multiple channels to screen for anomalous pulse signals, or they discriminate for background events by counting the number of residual pulses within a certain time after the pulse.

[0003] In recent years, the development of high-speed digitizers has made it possible to acquire the full waveform of fast liquid scintillator signals. During the acquisition process, the full waveform of each liquid scintillator pulse signal is collected and the data is sent to a computer via USB. Leveraging the powerful computing power of a computer, machine learning methods can be applied to further classify liquid scintillator pulse signals. Using machine learning methods such as feature dimensionality reduction, the full characteristics of each pulse can be effectively utilized, achieving better discrimination against background events.

[0004] Existing methods are effective to a certain extent, but they still have limitations. For example, while improvements in the construction materials of the detection system, such as increasing the thickness of the external lead shielding, can further reduce the background, this improvement has marginal effects and significantly increases the construction cost of the system. In addition, existing pulse discrimination methods based on signal processing have limited discrimination capabilities. In this process, when more relaxed parameters are selected, the background discrimination effect is limited. However, when stricter parameters are selected, more real events are mistakenly identified as background events, resulting in the loss of real events, thereby reducing the detection efficiency of the system. Ultimately, the ability to measure lower activity samples is not greatly improved. Summary of the Invention

[0005] The purpose of the present invention is to provide a background event screening method based on a random forest classification model.

[0006] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0007] The present invention comprises the following steps:

[0008] A high-activity standard sample and a blank sample are measured by a detection system to obtain an original data set, and the original data set is preprocessed to obtain a standard data set;

[0009] Constructing a random forest classification model according to the standard data set, training the random forest classification model, and optimizing the random forest classification model according to the training error; comprising:

[0010] The standard data set is reduced in dimension by principal component analysis, the characteristic matrix is ​​calculated, and the eigenvalues ​​are decomposed to obtain the eigenvalues ​​and eigenvectors. The expression is:

[0011]

[0012]

[0013] The standardized standard data matrix is ​​Z, and the transposed matrix of the standard data matrix Z is , the eigenvector matrix is , the eigenvector matrix The transposed matrix of , each column represents a eigenvector, and the diagonal matrix is , the elements on the diagonal are eigenvalues, the decomposition matrix is ​​S, and the number of eigenvalues ​​is n;

[0014] Extract eigenvalues ​​and eigenvectors according to the decomposition matrix, sort the eigenvalues ​​in descending order, select the first k largest eigenvalues ​​and corresponding eigenvectors, use the first k eigenvectors to construct a projection matrix, project the original data onto the principal component, and obtain the reduced dimensionality data. The expression is:

[0015]

[0016]

[0017] The kth eigenvector is , the first eigenvector is , the second eigenvector is , the projection matrix is ​​W, and the training data matrix is ​​Y;

[0018] Calculate the proportion of variance explained:

[0019]

[0020] The i-th explained variance ratio is , the i-th eigenvalue is , the number of standard data features is ;

[0021] The amount of information retained in the original data by the dimensionality reduction data was determined based on the explained variance ratio, and the principal component features with an explained variance greater than 95% were selected to establish a classification model based on random forests;

[0022] The optimized classification model is used to identify background events of the sample to be tested, and a time series relationship analysis is performed on the sample to be tested and the standard data set, and an identification result is output.

[0023] Furthermore, the method for training the random forest classification model includes:

[0024] The standard dataset after dimensionality reduction is split into a 4:1 ratio to obtain a training dataset and a validation dataset. The training dataset is used to train the classification model, and the validation dataset is used to evaluate the performance of the trained classification model.

[0025] During the classification model training process, the prediction results are integrated by building multiple decision trees. When building each decision tree, a random number of samples are extracted from the training data according to the principle of random sampling with replacement to form a new training set.

[0026] When constructing the nodes of the decision tree, k features are randomly selected from m features, and the number of selected features is calculated:

[0027]

[0028] The number of features is selected as k, the number of initial features is m, and the initial feature with the smallest Gini index is used as the splitting criterion to select the optimal splitting point. The expression of the Gini index is:

[0029]

[0030] The probability of category i is , the number of categories is , the Gini index is ;

[0031] Recursively select the best split point for the child nodes until the tree reaches the maximum depth, then stop the recursion;

[0032] The voting method is used to select the prediction result of the majority decision tree as the prediction result. The expression is:

[0033]

[0034] The prediction result of the T-th decision tree is , the prediction result of the first decision tree is , the prediction result of the second decision tree is , the number of decision trees is T.

[0035] Furthermore, the method for optimizing the random forest classification model according to the training error includes:

[0036] Obtain the training and validation datasets for the random forest classification model, and dynamically adjust the range of hyperparameters based on the size and feature dimensions of the training dataset.

[0037] Given the optimization objective function, the expression is:

[0038]

[0039]

[0040] The hyperparameter search space is , hyperparameter search space The objective function is , the total number of samples of training data is , the prediction result of the a-th training data is , the true label of the a-th training data is , the true label of the a-th training data in category i is , the prediction result of the a-th training data on category i is , the number of categories is , the loss function is , the weight coefficients are 、 、 , F1 score is F, accuracy is X;

[0041] The Gaussian process regression model based on the Gaussian kernel is used as the surrogate model. The acquisition function is used as the expectation according to the training data set, and the improved hyperparameters are used to obtain the optimal hyperparameter combination. The expression is:

[0042]

[0043]

[0044] The best hyperparameter combination is , the optimal hyperparameter combination includes the optimal number of decision trees and the optimal maximum depth. The hyperparameter search space is H, and the cth hyperparameter combination is , the jth hyperparameter combination is , hyperparameter combination and hyperparameter combinations The kernel function is , length scale is b, hyperparameter combination and hyperparameter combinations The standard deviation of ;

[0045] Based on the training dataset, the weighted sum of the F1 score and accuracy is used as the validation threshold for performance evaluation. A test dataset is selected to obtain a data subset. The performance evaluation value of the random forest classification model is calculated using the data subset. When the performance evaluation value is greater than the validation threshold, the validation threshold is updated; otherwise, negative feedback is triggered.

[0046] When negative feedback is triggered, a data subset is added to the training dataset and the hyperparameter search space is adjusted. The expression is:

[0047]

[0048]

[0049] The search space for the number of adjusted decision trees is , the adjusted maximum depth search space is , the index of the non-decreasing sorted number of decision trees is u, the index of the non-decreasing sorted maximum depth is a, the number of decision trees is e, and the maximum depth is , the number of adaptively adjusted decision trees is indexed as , the maximum depth index of adaptive adjustment is ;

[0050] Recalculate the current performance evaluation value. If it is greater than the validation threshold, traverse the test set and output the hyperparameter search space; otherwise, increase the number of data subsets and adjust the hyperparameter search space.

[0051] Furthermore, the method for performing time series relationship analysis on the sample to be tested and the standard data set includes:

[0052] Sort the samples to be tested and the standard dataset by time, use the samples to be tested as the experimental group and the standard dataset as the control group, and perform initial sliding window division on the experimental and control groups;

[0053] Calculate the mean and variance of the sliding data window data set of the experimental group and the control group, and standardize the window data;

[0054] A slow feature analysis model was established for the window, and feature mapping and dimensionality reduction were performed on the data of the experimental and control groups;

[0055] Perform correlation analysis on the slow feature analysis models of the experimental group and the control group, generate correlation spectra and clustering maps, and determine relevant data blocks based on the correlation spectra and clustering maps;

[0056] Convert the window data into data classification results. In the correlation map, if a large continuous window with a correlation greater than 0.759 is steady-state data, then re-establish the slow feature analysis model based on the steady-state data to obtain a steady-state slow feature analysis monitoring module. If the correlation is less than or equal to 0.759, set the short sliding window size, sliding step size, and correlation threshold.

[0057] Calculate the correlation between the short sliding window and the sliding window. If the correlation is greater than the correlation threshold, the short sliding window and the sliding window are correlated. Calculate the window merging length:

[0058]

[0059] The sliding step length is , the correlation between the short sliding window and the sliding window is , the window merging length is ;

[0060] Merge the first g data of the added window into the short sliding window, update the short sliding window to obtain the current window, and calculate the correlation between the current window and the sliding window. If the correlation is less than or equal to the correlation threshold, divide the short sliding window into sub-stages and establish a slow feature analysis model based on the sub-stages.

[0061] The unmodeled window data is divided according to the initial window length and the initial sliding step size to obtain data blocks. A synchronous propagation tree is established for the last s data of the first sub-stage and the previous steady-state data. The correlation rules between the experimental group and the control group are obtained based on the synchronous propagation tree.

[0062] A synchronous propagation tree is established for the last s data of the last sub-stage and the next steady-state data, and the correlation rules between the experimental group and the control group are obtained according to the synchronous propagation tree.

[0063] The beneficial effects of the present invention are:

[0064] The present invention is a background event identification method based on a random forest classification model. Compared with the existing technology, the present invention has the following technical effects:

[0065] This invention involves preprocessing, constructing a random forest classification model, training the random forest classification model, and optimizing the model. To improve the measurement capability of existing low-background liquid scintillation analyzers for even lower-activity samples without significantly increasing system construction costs, it is necessary to develop more effective background reduction strategies while maintaining detection efficiency. Through machine learning techniques, it is possible to more accurately distinguish between real signals and background events, thereby further reducing the system's background level and enhancing the system's ability to measure even lower-activity samples without significantly increasing system construction costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flowchart of the steps of a background event identification method based on a random forest classification model of the present invention;

[0067] Figure 2 A comparison chart of background energy spectra of the method according to the embodiment of this specification and that of the prior art. DETAILED DESCRIPTION

[0068] The present invention will be further described below through specific examples. The illustrative examples and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0069] The present invention provides a background event identification method based on a random forest classification model, comprising the following steps:

[0070] like Figure 1 As shown, in this embodiment, the following steps are included:

[0071] A high-activity standard sample and a blank sample are measured by a detection system to obtain an original data set, and the original data set is preprocessed to obtain a standard data set;

[0072] In the actual evaluation, a high-speed digital data acquisition instrument of a certain company with a sampling rate of 500M per second and model DT5730S was used as the experimental acquisition instrument;

[0073] Using a high-speed digital data acquisition instrument on a low-background liquid scintillation device, a standard 3H sample with an activity of 100 Bq and a blank sample were measured for 10 minutes, generating 5,000 labeled real event data and 5,000 labeled background event data sets. Data preprocessing was performed on 10,000 waveform data sets, and the final preprocessed data was used as the standard data set. The waveform data set: The original data set dimension is 10,000*360, with a total of 10,000 samples, each with 360 features.

[0074] Perform dimensionality reduction on the standard dataset and calculate the proportion of features that can be explained by each principal component after dimensionality reduction. It is found that the first 8 features can explain more than 95% of the variance. Finally, the first 8 features after dimensionality reduction are selected to form a new dataset. The dimension of the new dataset is 10000*8, with a total of 10,000 samples, and each sample has 8 features.

[0075] Constructing a random forest classification model according to the standard data set, training the random forest classification model, and optimizing the random forest classification model according to the training error; comprising:

[0076] The standard data set is reduced in dimension by principal component analysis, the characteristic matrix is ​​calculated, and the eigenvalues ​​are decomposed to obtain the eigenvalues ​​and eigenvectors. The expression is:

[0077]

[0078]

[0079] The standardized standard data matrix is ​​Z, and the transposed matrix of the standard data matrix Z is , the eigenvector matrix is , the eigenvector matrix The transposed matrix of , each column represents a eigenvector, and the diagonal matrix is , the elements on the diagonal are eigenvalues, the decomposition matrix is ​​S, and the number of eigenvalues ​​is n;

[0080] Extract eigenvalues ​​and eigenvectors according to the decomposition matrix, sort the eigenvalues ​​in descending order, select the first k largest eigenvalues ​​and corresponding eigenvectors, use the first k eigenvectors to construct a projection matrix, project the original data onto the principal component, and obtain the reduced dimensionality data. The expression is:

[0081]

[0082]

[0083] The kth eigenvector is , the first eigenvector is , the second eigenvector is , the projection matrix is ​​W, and the training data matrix is ​​Y;

[0084] Calculate the proportion of variance explained:

[0085]

[0086] The i-th explained variance ratio is , the i-th eigenvalue is , the number of standard data features is ;

[0087] The amount of information retained in the original data by the dimensionality reduction data was determined based on the explained variance ratio, and the principal component features with an explained variance greater than 95% were selected to establish a classification model based on random forests;

[0088] The optimized classification model is used to identify background events of the sample to be tested, and a time series relationship analysis is performed on the sample to be tested and the standard data set, and an identification result is output.

[0089] In this embodiment, the method for training the random forest classification model includes:

[0090] The standard dataset after dimensionality reduction is split into a 4:1 ratio to obtain a training dataset and a validation dataset. The training dataset is used to train the classification model, and the validation dataset is used to evaluate the performance of the trained classification model.

[0091] During the classification model training process, the prediction results are integrated by building multiple decision trees. When building each decision tree, a random number of samples are extracted from the training data according to the principle of random sampling with replacement to form a new training set.

[0092] When constructing the nodes of the decision tree, k features are randomly selected from m features, and the number of selected features is calculated:

[0093]

[0094] The number of features is selected as k, the number of initial features is m, and the initial feature with the smallest Gini index is used as the splitting criterion to select the optimal splitting point. The expression of the Gini index is:

[0095]

[0096] The probability of category i is , the number of categories is , the Gini index is ;

[0097] Recursively select the best split point for the child nodes until the tree reaches the maximum depth, then stop the recursion;

[0098] The voting method is used to select the prediction result of the majority decision tree as the prediction result. The expression is:

[0099]

[0100] The prediction result of the T-th decision tree is , the prediction result of the first decision tree is , the prediction result of the second decision tree is , the number of decision trees is T;

[0101] In the actual evaluation, the prediction results are integrated by building multiple decision trees. When building each decision tree, a random number of samples are randomly selected from the training data with replacement to form a new training set.

[0102] When constructing each node of the decision tree, 3 features are randomly selected from the 8 features, and the best split point is selected based on the selected features and the Gini index as the splitting criterion;

[0103] The above process is recursively repeated for each child node until the stopping condition is met, and the prediction result of the majority decision tree is selected as the prediction result using the voting method.

[0104] In this embodiment, the method for optimizing the random forest classification model according to the training error includes:

[0105] Obtain the training and validation datasets for the random forest classification model, and dynamically adjust the range of hyperparameters based on the size and feature dimensions of the training dataset.

[0106] Given the optimization objective function, the expression is:

[0107]

[0108]

[0109] The hyperparameter search space is , hyperparameter search space The objective function is , the total number of samples of training data is , the prediction result of the a-th training data is , the true label of the a-th training data is , the true label of the a-th training data in category i is , the prediction result of the a-th training data on category i is , the number of categories is , the loss function is , the weight coefficients are 、 、 , F1 score is F, accuracy is X;

[0110] The Gaussian process regression model based on the Gaussian kernel is used as the surrogate model. The acquisition function is used as the expectation according to the training data set, and the improved hyperparameters are used to obtain the optimal hyperparameter combination. The expression is:

[0111]

[0112]

[0113] The best hyperparameter combination is , the optimal hyperparameter combination includes the optimal number of decision trees and the optimal maximum depth. The hyperparameter search space is H, and the cth hyperparameter combination is , the jth hyperparameter combination is , hyperparameter combination and hyperparameter combinations The kernel function is , length scale is b, hyperparameter combination and hyperparameter combinations The standard deviation of ;

[0114] Based on the training dataset, the weighted sum of the F1 score and accuracy is used as the validation threshold for performance evaluation. A test dataset is selected to obtain a data subset. The performance evaluation value of the random forest classification model is calculated using the data subset. When the performance evaluation value is greater than the validation threshold, the validation threshold is updated; otherwise, negative feedback is triggered.

[0115] When negative feedback is triggered, a data subset is added to the training dataset and the hyperparameter search space is adjusted. The expression is:

[0116]

[0117]

[0118] The search space for the number of adjusted decision trees is , the adjusted maximum depth search space is , the index of the non-decreasing sorted number of decision trees is u, the index of the non-decreasing sorted maximum depth is a, the number of decision trees is e, and the maximum depth is , the number of adaptively adjusted decision trees is indexed as , the maximum depth index of adaptive adjustment is ;

[0119] Recalculate the current performance evaluation value. If it is greater than the validation threshold, traverse the test set and output the hyperparameter search space; otherwise, increase the number of data subsets and adjust the hyperparameter search space.

[0120] In the actual evaluation, the hyperparameters of the trained random forest classification model were optimized, mainly the number of decision trees and the number of node samples from 1 to 5. By trying all different parameter combinations, the optimal hyperparameters were obtained and the model with the best prediction effect on the validation dataset was solidified. The number of decision trees was increased from 100 to 1000, with a step size of 100.

[0121] The solidified random forest classification model was used to identify the background of the samples to be tested, and those identified as background events were eliminated.

[0122] In this embodiment, the method for performing time series relationship analysis on the sample to be tested and the standard data set includes:

[0123] Sort the samples to be tested and the standard dataset by time, use the samples to be tested as the experimental group and the standard dataset as the control group, and perform initial sliding window division on the experimental and control groups;

[0124] Calculate the mean and variance of the sliding data window data set of the experimental group and the control group, and standardize the window data;

[0125] A slow feature analysis model was established for the window, and feature mapping and dimensionality reduction were performed on the data of the experimental and control groups;

[0126] Perform correlation analysis on the slow feature analysis models of the experimental group and the control group, generate correlation spectra and clustering maps, and determine relevant data blocks based on the correlation spectra and clustering maps;

[0127] Convert the window data into data classification results. In the correlation map, if a large continuous window with a correlation greater than 0.759 is steady-state data, then re-establish the slow feature analysis model based on the steady-state data to obtain a steady-state slow feature analysis monitoring module. If the correlation is less than or equal to 0.759, set the short sliding window size, sliding step size, and correlation threshold.

[0128] Calculate the correlation between the short sliding window and the sliding window. If the correlation is greater than the correlation threshold, the short sliding window and the sliding window are correlated. Calculate the window merging length:

[0129]

[0130] The sliding step length is , the correlation between the short sliding window and the sliding window is , the window merging length is ;

[0131] Merge the first g data of the added window into the short sliding window, update the short sliding window to obtain the current window, and calculate the correlation between the current window and the sliding window. If the correlation is less than or equal to the correlation threshold, divide the short sliding window into sub-stages and establish a slow feature analysis model based on the sub-stages.

[0132] The unmodeled window data is divided according to the initial window length and the initial sliding step size to obtain data blocks. A synchronous propagation tree is established for the last s data of the first sub-stage and the previous steady-state data. The correlation rules between the experimental group and the control group are obtained based on the synchronous propagation tree.

[0133] The last s data of the last sub-stage and the next steady-state data are used to establish a synchronous propagation tree, and the correlation rules between the experimental group and the control group are obtained according to the synchronous propagation tree;

[0134] In the actual evaluation, the correlation threshold is 0.767.

[0135] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A background event identification method based on a random forest classification model, characterized in that: The following steps are involved: The detection system measures high-activity standard samples and blank samples to obtain a raw data set, and the raw data set is preprocessed to obtain a standard data set; the standard data set is measured on a low-background liquid scintillation device using a high-speed digital acquisition instrument for 10 minutes on a standard 3H sample with an activity of 100 Bq and a blank sample to obtain 5000 labeled real event data and 5000 labeled background event data sets; Constructing a random forest classification model according to the standard data set, training the random forest classification model, and optimizing the random forest classification model according to the training error; comprising: The standard data set is reduced in dimension by principal component analysis, the characteristic matrix is ​​calculated, and the eigenvalues ​​are decomposed to obtain the eigenvalues ​​and eigenvectors. The expression is: The standardized standard data matrix is ​​Z, and the transposed matrix of the standard data matrix Z is , the eigenvector matrix is , the eigenvector matrix The transposed matrix of , each column represents a eigenvector, and the diagonal matrix is , the elements on the diagonal are eigenvalues, the decomposition matrix is ​​S, and the number of eigenvalues ​​is n; Extract eigenvalues ​​and eigenvectors according to the decomposition matrix, sort the eigenvalues ​​in descending order, select the first k largest eigenvalues ​​and corresponding eigenvectors, use the first k eigenvectors to construct a projection matrix, project the original data onto the principal component, and obtain the reduced dimensionality data. The expression is: The kth eigenvector is , the first eigenvector is , the second eigenvector is , the projection matrix is ​​W, and the training data matrix is ​​Y; Calculate the proportion of variance explained: The i-th explained variance ratio is , the i-th eigenvalue is , the number of standard data features is ; The amount of information retained in the original data by the dimensionality reduction data was determined based on the explained variance ratio, and the principal component features with an explained variance greater than 95% were selected to establish a classification model based on random forests; include: Obtain the training and validation datasets for the random forest classification model, and dynamically adjust the range of hyperparameters based on the size and feature dimensions of the training dataset. Given the optimization objective function, the expression is: The hyperparameter search space is , hyperparameter search space The objective function is , the total number of samples of training data is , the prediction result of the a-th training data is , the true label of the a-th training data is , the true label of the a-th training data in category i is , the prediction result of the a-th training data on category i is , the number of categories is , the loss function is , the weight coefficients are 、 、 , F1 score is F, accuracy is X; The Gaussian process regression model based on the Gaussian kernel is used as the surrogate model. The acquisition function is used as the expectation according to the training data set, and the improved hyperparameters are used to obtain the optimal hyperparameter combination. The expression is: The best hyperparameter combination is , the optimal hyperparameter combination includes the optimal number of decision trees and the optimal maximum depth. The hyperparameter search space is H, and the cth hyperparameter combination is , the jth hyperparameter combination is , hyperparameter combination and hyperparameter combinations The kernel function is , length scale is b, hyperparameter combination and hyperparameter combinations The standard deviation of ; Based on the training dataset, the weighted sum of the F1 score and accuracy is used as the validation threshold for performance evaluation. A test dataset is selected to obtain a data subset. The performance evaluation value of the random forest classification model is calculated using the data subset. When the performance evaluation value is greater than the validation threshold, the validation threshold is updated; otherwise, negative feedback is triggered. When negative feedback is triggered, a data subset is added to the training dataset and the hyperparameter search space is adjusted. The expression is: The search space for the number of adjusted decision trees is , the adjusted maximum depth search space is , the index of the non-decreasing sorted number of decision trees is u, the index of the non-decreasing sorted maximum depth is a, the number of decision trees is e, and the maximum depth is , the number of adaptively adjusted decision trees is indexed as , the maximum depth index of adaptive adjustment is ; Recalculate the current performance evaluation value. If it is greater than the validation threshold, traverse the test set and output the hyperparameter search space; otherwise, increase the number of data subsets and adjust the hyperparameter search space. The optimized classification model is used to identify background events of the sample to be tested, and a time series relationship analysis is performed on the sample to be tested and the standard data set, and an identification result is output.

2. The background event identification method based on the random forest classification model according to claim 1, characterized in that: The method of training the random forest classification model comprises: The standard dataset after dimensionality reduction is split into a 4:1 ratio to obtain a training dataset and a validation dataset. The training dataset is used to train the classification model, and the validation dataset is used to evaluate the performance of the trained classification model. During the classification model training process, the prediction results are integrated by building multiple decision trees. When building each decision tree, a random number of samples are extracted from the training data according to the principle of random sampling with replacement to form a new training set. When constructing the nodes of the decision tree, k features are randomly selected from m features, and the number of selected features is calculated: The number of features is selected as k, the number of initial features is m, and the initial feature with the smallest Gini index is used as the splitting criterion to select the optimal splitting point. The expression of the Gini index is: The probability of category i is , the number of categories is , the Gini index is ; Recursively select the best split point for the child nodes until the tree reaches the maximum depth, then stop the recursion; The voting method is used to select the prediction result of the majority decision tree as the prediction result. The expression is: The prediction result of the T-th decision tree is , the prediction result of the first decision tree is , the prediction result of the second decision tree is , the number of decision trees is T.

3. The background event identification method based on the random forest classification model according to claim 1, characterized in that: The method for performing time series relationship analysis on the sample to be tested and the standard data set comprises: Sort the samples to be tested and the standard dataset by time, use the samples to be tested as the experimental group and the standard dataset as the control group, and perform initial sliding window division on the experimental and control groups; Calculate the mean and variance of the sliding data window data set of the experimental group and the control group, and standardize the window data; A slow feature analysis model was established for the window, and feature mapping and dimensionality reduction were performed on the data of the experimental and control groups; Perform correlation analysis on the slow feature analysis models of the experimental group and the control group, generate correlation spectra and clustering maps, and determine relevant data blocks based on the correlation spectra and clustering maps; Convert the window data into data classification results. In the correlation map, if a large continuous window with a correlation greater than 0.759 is steady-state data, then re-establish the slow feature analysis model based on the steady-state data to obtain a steady-state slow feature analysis monitoring module. If the correlation is less than or equal to 0.759, set the short sliding window size, sliding step size, and correlation threshold. Calculate the correlation between the short sliding window and the sliding window. If the correlation is greater than the correlation threshold, the short sliding window and the sliding window are correlated. Calculate the window merging length: The sliding step length is , the correlation between the short sliding window and the sliding window is , the window merging length is ; Merge the first g data of the added window into the short sliding window, update the short sliding window to obtain the current window, and calculate the correlation between the current window and the sliding window. If the correlation is less than or equal to the correlation threshold, divide the short sliding window into sub-stages and establish a slow feature analysis model based on the sub-stages. The unmodeled window data is divided according to the initial window length and the initial sliding step size to obtain data blocks. A synchronous propagation tree is established for the last s data of the first sub-stage and the previous steady-state data. The correlation rules between the experimental group and the control group are obtained based on the synchronous propagation tree. A synchronous propagation tree is established for the last s data of the last sub-stage and the next steady-state data, and the correlation rules between the experimental group and the control group are obtained according to the synchronous propagation tree.

Images