Wire and cable detection data refinement analysis and fault prediction system
By evaluating and empowering the decision tree in the random forest model, optimizing the wire and cable fault prediction results, the accuracy problem caused by the randomness of the decision tree in the wire and cable fault prediction is solved, and more accurate fault detection is achieved.
Patent Information
- Application Number
- CN202410630392.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-21
- Publication Date
- 2024-12-13
- Estimated Expiration
- 2044-05-21
AI Technical Summary
When using the random forest model to predict faults for wires and cables in the prior art, the construction of the decision tree is random, resulting in inaccurate final classification prediction results, affecting the accuracy of fault prediction.
The data processing module collects multi-dimensional detection timing data of wires and cables. The important evaluation module evaluates the degree of structural refinement and accuracy of the decision tree. The weight allocation module assigns weight values to each decision tree. The fault prediction module optimizes the final prediction results based on the weight values. , improve the accuracy of fault prediction.
Through the determination and optimization of adaptive important weights, the fault prediction results of wires and cables are more rigorous, and the fault detection is more accurate, which can better divide normal data and abnormal data.
Smart Images
Figure CN118364385B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of wire and cable detection, and in particular to a wire and cable detection data refinement analysis and fault prediction system. Background Art
[0002] The detection data of wires and cables usually contains characteristic data of multiple dimensions, such as current, voltage, temperature, resistance and other characteristic data. Since the collected characteristic data of each dimension have a certain correlation, if these characteristic data are analyzed in a single dimension, there are certain limitations, which will cause errors in the fault prediction results of wires and cables. Therefore, it is necessary to conduct detailed analysis and classification of the detection data of each dimension in the wire and cable scenario to ensure the accuracy of the fault prediction results of wires and cables.
[0003] Considering that the performance of wires and cables is affected by many factors, and there are complex nonlinear relationships between these factors in theory, random forest is an integrated learning algorithm that builds a powerful prediction model based on decision trees. The random forest model is composed of multiple decision trees, each of which is a classifier. The random forest model can adaptively capture the nonlinear relationship between features, that is, the random forest model can handle the noise and abnormalities in the multi-dimensional feature data of the scene, so as to establish a fault prediction model more accurately. Therefore, in the prior art, the trained random forest model is used to predict the faults of wires and cables based on the detection data of wires and cables.
[0004] However, when using the random forest model to predict faults of wires and cables, the final classification prediction result is obtained by voting or taking the average of the classification prediction results of all decision trees. In actual application, each decision tree is constructed by random sampling with replacement in the training set, which has a certain degree of randomness. The final classification prediction result obtained by voting or taking the average in the traditional way is inaccurate, that is, the prediction performance of the trained random forest model is poor, which leads to deviations in the fault classification prediction results of wires and cables.
[0005] Therefore, how to improve the accuracy of fault prediction of wires and cables using the trained random forest model has become an urgent problem to be solved. Summary of the invention
[0006] In view of this, an embodiment of the present invention provides a system for fine-grained analysis and fault prediction of wire and cable detection data to solve the problem of how to improve the accuracy of fault prediction of wires and cables using a trained random forest model.
[0007] An embodiment of the present invention provides a system for fine-grained analysis and fault prediction of wire and cable detection data, the system comprising:
[0008] A data processing module, used to collect detection time series data of at least two dimensions of the wire and cable, input the detection time series data of all dimensions into the trained random forest model, and obtain at least one decision tree;
[0009] An importance evaluation module is used to obtain, for any decision tree, the degree of structural refinement of the decision tree according to the depth of the decision tree and the number of leaf nodes on the decision tree, obtain the degree of structural accuracy of the decision tree according to the error characteristics and prediction accuracy of the out-of-bag samples of the decision tree, and obtain the importance of the decision tree in the trained random forest model according to the degree of structural refinement and structural accuracy of the decision tree;
[0010] A weight allocation module, used to obtain the importance of each decision tree in the trained random forest model, and obtain the weight value of each decision tree according to the importance of each decision tree in the trained random forest model;
[0011] A fault prediction module is used to obtain the final classification prediction result of the trained random forest model according to the classification prediction result and the corresponding weight value of each decision tree, and perform fault detection on the wires and cables according to the final classification prediction result.
[0012] Preferably, the important evaluation module obtains the structural refinement degree of the decision tree according to the depth of the decision tree and the number of leaf nodes on the decision tree, including:
[0013] Obtaining the depth of the decision tree according to the longest path length between a root node and a leaf node in the decision tree, and normalizing the depth of the decision tree to obtain a corresponding first normalized value;
[0014] Counting the number of leaf nodes included in the decision tree, normalizing the number of leaf nodes to obtain a corresponding second normalized value, and performing weighted summation on the first normalized value and the second normalized value to obtain a corresponding weighted summation result;
[0015] According to the number of samples contained in each leaf node in the decision tree, the minimum number of samples is obtained, the minimum number of samples is normalized to obtain a corresponding third normalized value, and the product of the third normalized value and the weighted sum result is used as the degree of structural refinement of the decision tree.
[0016] Preferably, the important evaluation module obtains the structural accuracy of the decision tree according to the error characteristics and prediction accuracy of the out-of-bag samples of the decision tree, including:
[0017] Obtaining an out-of-bag sample set of the decision tree, counting the total number of samples in the out-of-bag sample set and the number of samples misclassified in the out-of-bag sample set, calculating a first ratio between the total number of samples in the out-of-bag sample set and the number of samples misclassified in the out-of-bag sample set, substituting the opposite of the first ratio into an exponential function with a natural constant as the base, and obtaining a corresponding exponential function value;
[0018] The classification accuracy of the decision tree is obtained, the classification accuracy is normalized to obtain a corresponding normalized value, and a weighted sum of the normalized value and the exponential function value is used as the structural accuracy of the decision tree.
[0019] Preferably, the importance evaluation module obtains the importance of the decision tree in the trained random forest model according to the structural refinement and structural accuracy of the decision tree, including:
[0020] The structural refinement and structural accuracy of the decision tree are averaged, and the obtained average is used as the importance of the decision tree in the trained random forest model.
[0021] Preferably, the weight distribution module obtains the weight value of each decision tree according to the importance of each decision tree in the trained random forest model, including:
[0022] For any decision tree, the sum of a constant 1 and the importance of the decision tree in the trained random forest model is used as the weight value of the decision tree.
[0023] Preferably, the fault prediction module obtains the final classification prediction result of the trained random forest model according to the classification prediction result and the corresponding weight value of each decision tree, including:
[0024] According to the weight value of each decision tree, the classification prediction results of all decision trees are weighted, and the corresponding results are used as the final classification prediction results of the trained random forest model.
[0025] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:
[0026] The present invention provides a refined analysis and fault prediction system for wire and cable detection data, the system comprising: a data processing module, used to collect detection time series data of at least two dimensions of the wire and cable, input the detection time series data of all dimensions into a trained random forest model, and obtain at least one decision tree; an importance evaluation module, used to obtain the structural refinement degree of any decision tree according to the depth of the decision tree and the number of leaf nodes on the decision tree, obtain the structural accuracy of the decision tree according to the error characteristics and prediction accuracy of out-of-bag samples of the decision tree, and obtain the importance of the decision tree in the trained random forest model according to the structural refinement degree and structural accuracy of the decision tree; a weight allocation module, used to obtain the importance of each decision tree in the trained random forest model, and obtain the weight value of each decision tree according to the importance of each decision tree in the trained random forest model; a fault prediction module, used to obtain the final classification prediction result of the trained random forest model according to the classification prediction result of each decision tree and the corresponding weight value, and perform fault detection on the wire and cable according to the final classification prediction result. Among them, in the process of using the trained random forest model to predict faults of at least two dimensions of detection time series data of wires and cables, the prediction effect of each decision tree in the random forest model is analyzed to determine adaptive important weights, and the classification prediction results of the random forest model are optimized according to the determined weights, so that the division of normal data and abnormal data in the detection data of wires and cables is more in line with reality, thereby making the fault prediction results of wires and cables more rigorous and the fault detection more accurate. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0028] Figure 1 It is a structural block diagram of a wire and cable detection data refined analysis and fault prediction system provided in Example 1 of the present invention. DETAILED DESCRIPTION
[0029] Embodiments of the present disclosure are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.
[0030] It should be noted that the terms "first", "second", etc. in the specification of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure.
[0031] In order to illustrate the technical solution of the present invention, specific embodiments are provided below for illustration.
[0032] See also Figure 1 , is a structural block diagram of a wire and cable detection data refinement analysis and fault prediction system provided by the first embodiment of the present invention, such as Figure 1 As shown, the system may include:
[0033] The data processing module 11 is used to collect detection time series data of at least two dimensions of the wires and cables, input the detection time series data of all dimensions into the trained random forest model, and obtain at least one decision tree.
[0034] The detection data of the wires and cables include but are not limited to voltage, current, temperature, resistance, etc., and the embodiment of the present invention performs data analysis on the detection data in multiple dimensions. Therefore, the detection time series data of at least two dimensions of the wires and cables are collected.
[0035] In order to ensure the integrity and accuracy of the detection time series data of each dimension, the detection time series data of each dimension is first preprocessed by data cleaning and other preprocessing operations. The specific preprocessing operations are: check whether there are missing values or abnormal values in the collected detection time series data, and perform corresponding processing according to the inspection results. For example, for missing values, interpolation can be used to fill them to ensure the integrity of the detection time series data; for abnormal values (garbled values), they can be processed by limiting the range, replacing the average value or directly deleting them to ensure the accuracy of the detection time series data. It is worth noting that these preprocessing operations are all existing technologies and will not be described in detail here.
[0036] Random forest is a classifier that contains many decision trees. It can be used to handle classification and regression problems, as well as dimensionality reduction problems. It is also very tolerant to outliers and noise, and has better prediction and classification performance than decision trees. The Random Forest model is a classic bagging model, whose weak learner is a decision tree model. The Random Forest model randomly samples from the original data set to form n different sample data sets, and then builds n different decision tree models based on these data sets. Finally, the final result is obtained based on the average value (for regression models) or voting (for classification models) of these decision tree models.
[0037] Therefore, the collected detection time series data of all dimensions are used as a data set, and the data set is divided into a training set and a test set by cross-validation or retaining an independent test set. Then, the training set is used to train the random forest model to obtain a trained random forest model, which is used to predict the classification results of the detection data under each dimension. The training process of the random forest model belongs to the prior art and will not be repeated here. After obtaining the trained random forest model, the detection time series data of all dimensions are input into the trained random forest model, and multiple decision trees are obtained accordingly.
[0038] The importance evaluation module 12 is used to obtain, for any decision tree, the degree of structural refinement of the decision tree according to the depth of the decision tree and the number of leaf nodes on the decision tree, obtain the structural accuracy of the decision tree according to the error characteristics and prediction accuracy of the out-of-bag samples of the decision tree, and obtain the importance of the decision tree in the trained random forest model according to the degree of structural refinement and structural accuracy of the decision tree.
[0039] Since the random forest model obtains the final classification prediction result by voting or averaging the classification prediction results of all decision trees, and considering that decision trees with different structural characteristics should have different contribution ratios in the final classification prediction result, and the detection data of wires and cables needs to be refined for classification analysis in the scenario, the embodiment of the present invention analyzes the tree structure of each decision tree to determine the importance of each decision tree in the trained random forest model, so as to adaptively assign corresponding weight values.
[0040] Taking a decision tree as an example, according to the tree structure characteristics and classification accuracy of the decision tree, the importance of the decision tree in the trained random forest model is obtained. The specific process is as follows:
[0041] (1) Obtaining the degree of structural refinement of the decision tree according to the depth of the decision tree and the number of leaf nodes on the decision tree.
[0042] Specifically, according to the longest path length between the root node and the leaf nodes in the decision tree, the depth of the decision tree is obtained, and the depth of the decision tree is normalized to obtain a corresponding first normalized value;
[0043] Counting the number of leaf nodes included in the decision tree, normalizing the number of leaf nodes to obtain a corresponding second normalized value, and performing weighted summation on the first normalized value and the second normalized value to obtain a corresponding weighted summation result;
[0044] According to the number of samples contained in each leaf node in the decision tree, the minimum number of samples is obtained, the minimum number of samples is normalized to obtain a corresponding third normalized value, and the product of the third normalized value and the weighted sum result is used as the degree of structural refinement of the decision tree.
[0045] In one embodiment, the calculation expression of the structural refinement degree of the decision tree is:
[0046]
[0047] in, represents the degree of structural refinement of the I-th decision tree, Norm() represents the normalization function, represents the minimum number of samples contained in the leaf nodes of the I-th decision tree, represents the depth of the I-th decision tree, represents the number of leaf nodes in the I-th decision tree, represents the first weight, Represents the second weight.
[0048] Preferably, the reference values are given in the embodiments of the present invention =0.6, =0.4.
[0049] It should be noted that the greater the depth of the decision tree (the length of the longest path from the root node to the leaf node), the higher the complexity of the current decision tree. At this time, a more refined classification division of the detection time series data of all dimensions can be achieved; at the same time, the number of leaf nodes of the decision tree is proportional to the complexity of the decision tree. The more leaf nodes there are, the more decision paths and divisions there are. At this time, the higher the complexity of the decision tree, the The value and The larger the value, the more refined the structure of the corresponding decision tree is; on the other hand, the minimum number of samples contained in each leaf node of the decision tree can be used to evaluate the classification effect of the current decision tree. If the minimum number of samples is too small, it means that the current decision tree may have an overfitting problem, which is inversely proportional to the partitioning effect of the decision tree in theory. Therefore, The smaller the value, the less refined the structure of the corresponding decision tree, and the minimum number of samples contained in the leaf node is relatively large, which can avoid overfitting of the decision tree and ensure a certain decision tree generation effect.
[0050] (2) Obtaining the structural accuracy of the decision tree based on the error characteristics and prediction accuracy of the out-of-bag samples of the decision tree.
[0051] Specifically, obtaining an out-of-bag sample set of the decision tree, counting the total number of samples in the out-of-bag sample set and the number of samples misclassified in the out-of-bag sample set, calculating a first ratio between the total number of samples in the out-of-bag sample set and the number of samples misclassified in the out-of-bag sample set, substituting the opposite of the first ratio into an exponential function with a natural constant as the base, and obtaining a corresponding exponential function value;
[0052] The classification accuracy of the decision tree is obtained, the classification accuracy is normalized to obtain a corresponding normalized value, and a weighted sum of the normalized value and the exponential function value is used as the structural accuracy of the decision tree.
[0053] In one embodiment, the calculation expression of the structural accuracy of the decision tree is:
[0054]
[0055] in, represents the structural accuracy of the I-th decision tree, exp() represents an exponential function with a natural constant as the base, B represents the number of misclassified samples in the out-of-bag sample set of the I-th decision tree, represents the total number of samples in the out-of-bag sample set of the I-th decision tree, represents any sample in the out-of-bag sample set of the I-th decision tree, Norm() represents the normalization function, represents the number of samples correctly classified by the I-th decision tree in the test set, A represents the number of samples in the test set, Indicates the first The classification prediction result of samples under the I-th decision tree, Indicates the first The true category or label to which the samples belong, represents the third weight, Represents the fourth weight.
[0056] Preferably, the reference values are given in the embodiments of the present invention =0.7, =0.3.
[0057] It should be noted that The value of is used to characterize the error value of the out-of-bag samples of the current decision tree, that is, when constructing each decision tree, samples are extracted from the training set by sampling with replacement. For each sample, there is a certain probability that it will not be selected when it is extracted. These unselected samples are the out-of-bag samples of the current decision tree. It is worth noting that out-of-bag samples can be obtained when constructing a decision tree, which belongs to the prior art and will not be described here. For the current decision tree, unselected samples (i.e., out-of-bag samples) are used for prediction. This type of sample is not used for training when constructing the decision tree, so it can be used to estimate the performance of the current decision tree on unseen data. Therefore, this error can be used as one of the criteria for judging the structural accuracy of the current decision tree, and is used to evaluate the generalization ability of the current decision tree. Therefore, The smaller the value, the smaller the error, and the greater the accuracy of the current decision tree structure; at the same time, The value of is used to characterize the accuracy of the current decision tree in classifying and predicting the test set, that is, the proportion of correctly classified samples. By comparing the classification prediction results of the decision tree for the test set with the true labels of the samples in the test set, the number of correctly classified samples is counted and divided by the total number to obtain the accuracy calculation result of the test data set. It reflects the accuracy of the current decision tree in classifying known data. Therefore, The larger the value, the higher the accuracy of the current decision tree structure.
[0058] (3) Obtaining the importance of the decision tree in the trained random forest model based on the structural refinement and structural accuracy of the decision tree.
[0059] Specifically, the structural refinement and structural accuracy of the decision tree are averaged, and the obtained average is used as the importance of the decision tree in the trained random forest model.
[0060] In one embodiment, the calculation expression of the importance of the decision tree in the trained random forest model is:
[0061]
[0062] in, It indicates the importance of the I-th decision tree in the trained random forest model, AVG() indicates the mean function, Indicates the structural accuracy of the I-th decision tree, Indicates the degree of structural refinement of the I-th decision tree.
[0063] It should be noted that the higher the structural accuracy and refinement of the I-th decision tree, the stronger the prediction ability of the I-th decision tree, the smaller the error, and the higher its importance in the trained random forest model.
[0064] At this point, we can get the importance of the I-th decision tree in the trained random forest model.
[0065] The weight allocation module 13 is used to obtain the importance of each decision tree in the trained random forest model, and obtain the weight value of each decision tree according to the importance of each decision tree in the trained random forest model.
[0066] In the importance evaluation module 12, the importance of each decision tree in the trained random forest model can be obtained. After obtaining the importance of each decision tree in the trained random forest model, the importance is sorted, and a weight value is assigned to the corresponding decision tree according to the importance. The higher the importance, the larger the corresponding weight value.
[0067] Preferably, the weight value of each decision tree is obtained according to the importance of each decision tree in the trained random forest model, including: for any decision tree, taking the sum of a constant 1 and the importance of the decision tree in the trained random forest model as the weight value of the decision tree.
[0068] In one embodiment, the calculation expression of the weight value of the decision tree is: , where 1 represents a constant, represents the weight value of the I-th decision tree, Indicates the importance of the I-th decision tree in the trained random forest model.
[0069] At this point, the weight value of each decision tree can be obtained.
[0070] The fault prediction module 14 is used to obtain the final classification prediction result of the trained random forest model according to the classification prediction result and the corresponding weight value of each decision tree, and perform fault detection on the wires and cables according to the final classification prediction result.
[0071] After obtaining the weight value of each decision tree, the classification prediction results corresponding to the output of the trained random forest model are optimized according to the weight value of each decision tree. The specific process is as follows: first, the detection time series data of all dimensions are input into the trained random forest model to obtain the classification prediction results of each decision tree, and then the final classification prediction results of the trained random forest model are obtained according to the classification prediction results of each decision tree and the corresponding weight value. It is worth noting that the classification prediction results are also the classification results of normal data and abnormal data for the detection time series data of each dimension.
[0072] Preferably, according to the classification prediction results and corresponding weight values of each decision tree, the final classification prediction results of the trained random forest model are obtained, including:
[0073] According to the weight value of each decision tree, the classification prediction results of all decision trees are weighted, and the corresponding results are used as the final classification prediction results of the trained random forest model.
[0074] In one embodiment, the classification prediction results of each decision tree are weighted according to the weight value of each decision tree. If the weight values of each decision tree are equal, the classification prediction results of each decision tree can be simply summed or averaged. However, if the weight values of each decision tree are different, it is necessary to multiply the classification prediction results of each decision tree by the corresponding weight value "1+R(I)" to obtain the weighted classification prediction results of each decision tree. According to the weighted classification prediction results, the final classification prediction results are determined using appropriate voting rules. For example, for classification problems, a majority voting rule can be used. For example, a decision tree with a large weight value can be considered as two votes for one vote, that is, the category with the most votes is selected as the final classification prediction result; for regression problems, the average value of all weighted classification prediction results can be simply used as the final classification prediction result.
[0075] After obtaining the final classification prediction results of the trained random forest model, fault detection of wires and cables can be performed according to the final classification prediction results to provide early fault warning. It is worth noting that the focus of the present invention is on how to obtain the classification prediction results of classifying the detection time series data of all dimensions of wires and cables using the random forest model, and fault detection of wires and cables according to the final classification prediction results is not the focus of the present invention and will not be described in detail here.
[0076] In summary, an embodiment of the present invention provides a refined analysis and fault prediction system for wire and cable detection data, the system comprising: a data processing module, used to collect detection time series data of at least two dimensions of the wire and cable, input the detection time series data of all dimensions into a trained random forest model, and obtain at least one decision tree; an importance evaluation module, used to obtain the structural refinement of any decision tree according to the depth of the decision tree and the number of leaf nodes on the decision tree, obtain the structural accuracy of the decision tree according to the error characteristics and prediction accuracy of the out-of-bag samples of the decision tree, and obtain the importance of the decision tree in the trained random forest model according to the structural refinement and structural accuracy of the decision tree; a weight allocation module, used to obtain the importance of each decision tree in the trained random forest model, and obtain the weight value of each decision tree according to the importance of each decision tree in the trained random forest model; a fault prediction module, used to obtain the final classification prediction result of the trained random forest model according to the classification prediction result of each decision tree and the corresponding weight value, and perform fault detection on the wire and cable according to the final classification prediction result. Among them, in the process of using the trained random forest model to predict faults of at least two dimensions of detection time series data of wires and cables, the prediction effect of each decision tree in the random forest model is analyzed to determine the adaptive important weights, and the classification prediction results of the random forest model are optimized according to the determined weights, so that the fault classification prediction results of wires and cables are more rigorous and the fault detection is more accurate.
[0077] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.
Claims
1. A wire and cable detection data refinement analysis and fault prediction system, characterized in that: The system comprises: A data processing module, used to collect detection time series data of at least two dimensions of the wire and cable, input the detection time series data of all dimensions into the trained random forest model, and obtain at least one decision tree; An importance evaluation module is used to obtain, for any decision tree, the degree of structural refinement of the decision tree according to the depth of the decision tree and the number of leaf nodes on the decision tree, obtain the degree of structural accuracy of the decision tree according to the error characteristics and prediction accuracy of the out-of-bag samples of the decision tree, and obtain the importance of the decision tree in the trained random forest model according to the degree of structural refinement and structural accuracy of the decision tree; A weight allocation module, used to obtain the importance of each decision tree in the trained random forest model, and obtain the weight value of each decision tree according to the importance of each decision tree in the trained random forest model; A fault prediction module, used to obtain the final classification prediction result of the trained random forest model according to the classification prediction result and the corresponding weight value of each decision tree, and perform fault detection on the wire and cable according to the final classification prediction result; The important evaluation module obtains the structural refinement degree of the decision tree according to the depth of the decision tree and the number of leaf nodes on the decision tree, including: Obtaining the depth of the decision tree according to the longest path length between a root node and a leaf node in the decision tree, and normalizing the depth of the decision tree to obtain a corresponding first normalized value; Counting the number of leaf nodes included in the decision tree, normalizing the number of leaf nodes to obtain a corresponding second normalized value, and performing weighted summation on the first normalized value and the second normalized value to obtain a corresponding weighted summation result; According to the number of samples contained in each leaf node in the decision tree, a minimum number of samples is obtained, the minimum number of samples is normalized to obtain a corresponding third normalized value, and the product of the third normalized value and the weighted sum result is used as the structural refinement degree of the decision tree; The important evaluation module obtains the structural accuracy of the decision tree according to the error characteristics and prediction accuracy of the out-of-bag samples of the decision tree, including: Obtaining an out-of-bag sample set of the decision tree, counting the total number of samples in the out-of-bag sample set and the number of samples misclassified in the out-of-bag sample set, calculating a first ratio between the total number of samples in the out-of-bag sample set and the number of samples misclassified in the out-of-bag sample set, substituting the opposite of the first ratio into an exponential function with a natural constant as the base, and obtaining a corresponding exponential function value; Obtaining the classification accuracy of the decision tree, normalizing the classification accuracy to obtain a corresponding normalized value, and taking a weighted sum of the normalized value and the exponential function value as the structural accuracy of the decision tree; The weight distribution module obtains the weight value of each decision tree according to the importance of each decision tree in the trained random forest model, including: For any decision tree, the sum of a constant 1 and the importance of the decision tree in the trained random forest model is used as the weight value of the decision tree.
2. The wire and cable detection data refinement analysis and fault prediction system according to claim 1 is characterized in that: The importance evaluation module obtains the importance of the decision tree in the trained random forest model according to the structural refinement and structural accuracy of the decision tree, including: The structural refinement and structural accuracy of the decision tree are averaged, and the obtained average is used as the importance of the decision tree in the trained random forest model.
3. The wire and cable detection data refinement analysis and fault prediction system according to claim 1 is characterized in that: The fault prediction module obtains the final classification prediction result of the trained random forest model according to the classification prediction result of each decision tree and the corresponding weight value, including: According to the weight value of each decision tree, the classification prediction results of all decision trees are weighted, and the corresponding results are used as the final classification prediction results of the trained random forest model.