Railway data asset management system

Abstract

The invention provides a railway data asset management system, which relates to the technical field of data management, and comprises the steps of receiving railway operation data and maintenance records, performing data preprocessing to extract key information, and using the preprocessed data to generate a plurality of multi-dimensional feature vectors; sampling the multiple multi-dimensional feature vectors according to multiple randomly generated dimensions to obtain multiple groups of feature sub-vectors, and performing correlation analysis on the multiple groups of feature sub-vectors to obtain multiple groups of correlation results; obtaining a symmetric correlation matrix according to the attribute value difference of any two groups of feature sub-vectors in the plurality of groups of feature sub-vectors, calculating the weight of the plurality of groups of correlation results according to the symmetric correlation matrix, and matching counting judgment conditions according to the weight and the railway data to obtain a plurality of groups of correlation results. And calculating the weighted matching degree of each class of each data in the multiple groups of correlation results, and taking the class corresponding to the maximum weighted matching degree as the optimal matching class of the data.

Description

Technical Field

[0001] The present invention relates to the technical field of data management, and more specifically, to a railway data asset management system. Background Art

[0002] Existing railway data asset management systems mainly rely on traditional data processing and analysis methods, which are often inefficient and inaccurate when processing large-scale, multi-dimensional railway operation data and maintenance records. These systems usually lack the ability to mine and analyze deep-level data features, resulting in failure to achieve ideal results when matching and classifying data. In terms of technical principles, traditional systems mostly use simple statistical analysis and linear models, which make it difficult to capture complex correlations and nonlinear relationships between data, which limits the performance and reliability of the system when processing railway data.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: the existing railway data asset management system is often unable to effectively extract key information from a large amount of complex data when processing and analyzing railway operation data, resulting in low efficiency and accuracy of data preprocessing; at the same time, due to the lack of in-depth analysis of multi-dimensional feature vectors, the system is difficult to achieve high-precision recognition when performing data matching and classification, which directly affects the quality and efficiency of railway data asset management. In addition, the existing system also has deficiencies in correlation analysis and weight calculation, and cannot fully utilize the correlation information between data to optimize the matching results. These problems limit the effectiveness and application scope of the railway data asset management system. Summary of the invention

[0004] The present invention provides a railway data asset management system, comprising: receiving railway operation data and maintenance records, performing data preprocessing to extract key information, and using the preprocessed data to generate multiple multi-dimensional feature vectors; Sampling the plurality of multidimensional feature vectors according to the number of dimensions randomly generated multiple times to obtain a plurality of groups of feature sub-vectors, and performing correlation analysis on the plurality of groups of feature sub-vectors respectively to obtain a plurality of groups of correlation results; According to the difference in attribute values ​​between any two groups of feature sub-vectors in the multiple groups of feature sub-vectors, a symmetric correlation matrix is ​​obtained. According to the symmetric correlation matrix, the weights of the multiple groups of correlation results are calculated. According to the weights and the railway data matching count judgment condition, the weighted matching degree of each category of each data in the multiple groups of correlation results is calculated, and the category corresponding to the maximum weighted matching degree is taken as the optimal matching category of the data.

[0005] Further, the following steps are included: Performing statistical analysis on the dimensions of the correlation results to determine the distribution range of key dimensions; Based on the importance and distribution range of the key dimensions, 70% to 80% of the key dimensions are selected as the number of dimensions, which is used as the basis for sampling; On the basis of the benchmark, multiple values ​​are selected by random sampling method, the multiple values ​​are used as the number of dimensions, the number of values ​​is used as the sampling times, and the sampling times is no less than 3 times.

[0006] Furthermore, the sampling of the plurality of multidimensional feature vectors according to the number of dimensions randomly generated for a plurality of times to obtain a plurality of groups of feature sub-vectors includes: According to the dimension number ci obtained by the i-th sampling, a non-repeating number ci is randomly generated within the dimension range of the multidimensional feature vector, and after being arranged from small to large, the eigenvalues ​​of the corresponding dimensional sequence position are extracted for each multidimensional feature vector as a dimensional sequence to obtain the i-th group of feature sub-vectors; The feature sub-vectors obtained by each sampling constitute multiple groups of feature sub-vectors.

[0007] Furthermore, the correlation analysis uses an improved Pearson correlation coefficient algorithm to calculate two initial correlation centers according to the attribute value of each dimension in the feature subvector. The formula is: in, and are the attribute values ​​of the rth dimension in the feature sub-vector, and They are and The average value of is the total number of dimensions, r is the correlation coefficient between X and Y, X is the attribute value of the Xth group of feature sub-vectors, and Y is the attribute value of the Yth group of feature sub-vectors.

[0008] Furthermore, the symmetric correlation matrix is ​​obtained according to the attribute value difference between any two groups of feature sub-vectors in the multiple groups of feature sub-vectors, including: The Jaccard coefficient similarity of any two sets of feature sub-vectors is calculated by the following formula: in, and are the dimension sequences of any two sets of feature sub-vectors, for and The number of dimensions contained in for Contains, and The number of dimensions not included, for Contains, and The number of dimensions not included; Based on the Jaccard coefficient similarity of the two sets of feature sub-vectors, the dissimilarity of the two sets of feature sub-vectors is calculated using the following formula: : Based on the dissimilarity Get the symmetric correlation matrix : The values ​​of the diagonal elements are 0.

[0009] Furthermore, the step of calculating the weights of the plurality of groups of correlation results according to the symmetric correlation matrix includes: Using the following formula, multiple first weights are obtained according to the ratio of the sum of the element values ​​of each row in the matrix to the sum of all element values: The obtained multiple first weights are normalized and used as weights of multiple groups of correlation results.

[0010] Furthermore, the railway data matching count judgment condition is expressed as: in, is the data in the i-th group of correlation results , when the data If it is matched to the kth category, the count is 1, otherwise the count is 0.

[0011] Furthermore, according to the weight and the railway data matching count judgment condition, the weighted matching degree of each category of each data in the multiple groups of correlation results is calculated, and the calculation formula is: in, For data The weighted matching degree of matching it to the kth category in the R group correlation results.

[0012] Furthermore, in the data preprocessing, the number of nodes in the input layer is the number of dimensions of the railway data after preprocessing, the number of nodes in the output layer is the number of categories to which the railway data belongs, and the number of nodes in the last layer of the middle layer of the network is less than the number of dimensions of the input and greater than the number of categories of the output.

[0013] Furthermore, the preprocessing includes: extracting key information fragments from the received railway data; Using principal component analysis PCA to reduce the dimension of the key information fragments; The preprocessed railway data is obtained by normalizing the reduced-dimensional railway data.

[0014] The above-mentioned embodiments of the present invention have at least the following beneficial effects: the railway data asset management system of the present invention can significantly improve the processing efficiency and accuracy of railway operation data and maintenance records by adopting advanced data preprocessing technology and multi-dimensional feature vector analysis methods. The system receives railway operation data and maintenance records, performs data preprocessing to extract key information, and generates multiple multi-dimensional feature vectors, and then samples these feature vectors through multiple randomly generated dimensions to obtain multiple groups of feature sub-vectors, and performs correlation analysis. This method can deeply explore the complex correlations between data and provide a more accurate basis for railway data matching.

[0015] In addition, the system can effectively determine the optimal matching category of data by calculating the symmetric correlation matrix and weighted matching degree. This method based on weight and matching count judgment conditions can optimize the data matching process and improve the accuracy and reliability of matching. At the same time, the system also uses an improved Pearson correlation coefficient algorithm and Jaccard coefficient similarity calculation to further improve the accuracy of correlation analysis. The integrated application of these technologies can not only enhance the intelligent level of railway data asset management, but also provide more scientific and efficient data support for railway operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which: Figure 1 A schematic diagram of a flow chart of a railway data asset management system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0018] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, device, apparatus, method or computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0019] It should be noted that any number of elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning.

[0020] Reference below Figure 1 , Figure 1 The following is a flow chart of a railway data asset management system provided by an embodiment of the present invention. Figure 1 As shown, a railway data asset management system 100 includes: Step 101, receiving railway operation data and maintenance records, performing data preprocessing to extract key information, and using the preprocessed data to generate multiple multi-dimensional feature vectors; Step 102, sampling the plurality of multidimensional feature vectors according to the number of dimensions randomly generated multiple times to obtain a plurality of groups of feature sub-vectors, and performing correlation analysis on the plurality of groups of feature sub-vectors respectively to obtain a plurality of groups of correlation results; Step 103, based on the attribute value difference between any two groups of feature sub-vectors in the multiple groups of feature sub-vectors, a symmetric correlation matrix is ​​obtained, and based on the symmetric correlation matrix, the weights of the multiple groups of correlation results are calculated, and based on the weights and the railway data matching count judgment condition, the weighted matching degree of each category of each data in the multiple groups of correlation results is calculated, and the category corresponding to the maximum weighted matching degree is taken as the optimal matching category of the data.

[0021] It should be noted that the system receives railway operation data and maintenance records and performs data preprocessing to extract key information. Data preprocessing here refers to the process of cleaning, formatting and converting raw data in order to extract useful information from it. Key information may include train operation status, maintenance logs, fault records, etc., which are crucial for railway data asset management.

[0022] Specifically, the data preprocessing step includes cleaning the received railway data to remove invalid or erroneous data points; then formatting to ensure that the data follows a unified standard; and finally converting to extract key information fragments that are important for railway operation and maintenance. For example, key parameters such as train speed and arrival time can be extracted from the train operation log, and these parameters can be normalized for subsequent analysis and processing.

[0023] Preferably, data preprocessing may also include using principal component analysis (PCA) to reduce the dimensionality of key information fragments to reduce the complexity of the data and extract the most important features. In addition, the reduced-dimensional data is normalized to ensure that the comparison between different eigenvalues ​​is fair, which can improve the accuracy of subsequent correlation analysis. For example, eigenvalues ​​such as speed and time can be scaled to the range of [0,1] for effective comparison and analysis.

[0024] In some embodiments, the steps include: Performing statistical analysis on the dimensions of the correlation results to determine the distribution range of key dimensions; Based on the importance and distribution range of the key dimensions, 70% to 80% of the key dimensions are selected as the number of dimensions, which is used as the basis for sampling; On the basis of the benchmark, multiple values ​​are selected by random sampling method, the multiple values ​​are used as the number of dimensions, the number of values ​​is used as the sampling times, and the sampling times is no less than 3 times.

[0025] It should be noted that the system performs statistical analysis on the dimensions of the output results to determine the distribution range of key dimensions. Statistical analysis here refers to the use of statistical methods to analyze data to identify which dimensions are most important in the data set. Key dimensions refer to those data features that have a significant impact on railway operation and maintenance decisions. The distribution range refers to the range of values ​​of these key dimensions in the data set.

[0026] Specifically, the system determines which dimensions are critical by analyzing the frequency, mean, variance and other statistical indicators of each dimension in the data set. For example, the amount of information can be evaluated by calculating the variance of each dimension. The larger the variance, the more dispersed the data distribution of the dimension is, and it may contain more useful information. In addition, the system will also consider the correlation between dimensions to avoid selecting highly correlated dimensions, thereby reducing data redundancy.

[0027] Preferably, the system selects 70% to 80% of the key dimensions as the sampling basis based on their importance and distribution range. This means that the system will give priority to the most representative and informative dimensions for subsequent analysis. For example, if a data set has 10 dimensions, the system may select 7 to 8 of the most important dimensions as the basis for analysis.

[0028] Furthermore, the system can use random sampling methods in the sampling process to ensure the diversity and representativeness of the sample. In actual operation, the system may set a random seed to ensure the repeatability of sampling, or use techniques such as stratified sampling to ensure that key dimensions are fully considered.

[0029] In some embodiments, sampling the plurality of multidimensional feature vectors according to the number of dimensions randomly generated multiple times to obtain a plurality of groups of feature sub-vectors includes: According to the dimension number ci obtained by the i-th sampling, a non-repeating number ci is randomly generated within the dimension range of the multidimensional feature vector, and after being arranged from small to large, the eigenvalues ​​of the corresponding dimensional sequence position are extracted for each multidimensional feature vector as a dimensional sequence to obtain the i-th group of feature sub-vectors; The feature sub-vectors obtained by each sampling constitute multiple groups of feature sub-vectors.

[0030] It should be noted that the system samples multiple multidimensional feature vectors according to the number of randomly generated dimensions to obtain multiple groups of feature sub-vectors. The randomly generated number of dimensions here means that the system does not select the same dimension each time it samples, but randomly selects a certain number of dimensions. This can increase the diversity of samples and improve the robustness of the analysis results. A multidimensional feature vector refers to a data point that contains multiple features, and each feature corresponds to a dimension in the vector. A feature sub-vector refers to a vector composed of a part of the dimensions extracted from a multidimensional feature vector.

[0031] Specifically, when the system samples for the i-th time, it will randomly generate a non-repeating number of ci within the dimensional range of the multidimensional feature vector based on the number of dimensions ci obtained by sampling, and arrange them from small to large as a dimensional sequence. Then, for each multidimensional feature vector, the eigenvalues ​​corresponding to the dimensional sequence position are extracted to obtain the i-th group of feature sub-vectors. This process will be repeated many times, and the number of dimensions obtained by each sampling may be different, thus forming multiple groups of feature sub-vectors. For example, if a multidimensional feature vector has 10 dimensions, the system may select dimensions 3, 5, and 7 in the first sampling, and dimensions 1, 4, and 9 in the second sampling. Each extracted eigenvalue will form a new feature sub-vector.

[0032] Preferably, when performing random sampling, the system can adopt different random number generation strategies, such as a pseudo-random number generator or a random sampling method based on a certain distribution, to ensure that the number of dimensions sampled each time is both random and representative.

[0033] Furthermore, the system can also set a lower limit for the number of sampling times, such as no less than 3 times, to ensure that enough samples can be obtained for correlation analysis. In actual operation, the system may record the dimension sequence and corresponding feature sub-vector of each sampling for subsequent analysis and comparison.

[0034] In some embodiments, the correlation analysis uses an improved Pearson correlation coefficient algorithm to calculate two initial correlation centers based on the attribute value of each dimension in the feature subvector. The formula is: in, and are the attribute values ​​of the rth dimension in the feature sub-vector, and They are and The average value of is the total number of dimensions. It should be noted that the system uses the improved Pearson correlation coefficient algorithm to calculate two initial correlation centers for the attribute values ​​of each dimension in the feature sub-vector. The Pearson correlation coefficient is a statistical indicator that measures the degree of linear correlation between two variables. Its value is between -1 and 1. The closer the value is to 1 or -1, the stronger the correlation is, and the closer the value is to 0, the weaker the correlation is. The improved Pearson correlation coefficient algorithm refers to the optimization of the traditional algorithm to adapt to specific data characteristics or improve calculation efficiency. The initial correlation center refers to the benchmark value used to compare and calculate correlation in correlation analysis.

[0035] Specifically, the system calculates the sum of the deviations between the attribute value of the rth dimension in the feature sub-vector and the average value through a formula, and then divides it by the square root of the product of the sum of the squares of the attribute value deviations of the two dimensions to obtain the Pearson correlation coefficient. The variables involved in the formula here include the attribute value of the rth dimension in the feature sub-vector ( and ), and their averages ( and ). In this way, the system can quantify the linear correlation between two dimensions. For example, if and Representing the speed and running time of the train respectively, the system will calculate the correlation coefficient of these two dimensions to evaluate the strength of the correlation between them.

[0036] Preferably, when calculating the Pearson correlation coefficient, the system can standardize the data to eliminate the influence of different dimensions. In addition, the system can also introduce weight factors to assign different weights to data of different dimensions to reflect their importance in railway operations. For example, if the punctuality of a train is more important than its speed, the system can assign a higher weight to punctuality. In actual operation, the system may use ready-made functions in the software library to calculate the Pearson correlation coefficient, or implement the algorithm itself as needed.

[0037] In some embodiments, obtaining a symmetric correlation matrix according to the attribute value difference between any two groups of feature sub-vectors in the plurality of groups of feature sub-vectors includes: calculating the Jaccard coefficient similarity between any two groups of feature sub-vectors by the following formula: in, and are the dimension sequences of any two sets of feature sub-vectors, for and The number of dimensions contained in for Contains, and The number of dimensions not included, for Contains, and The number of dimensions not included; Based on the Jaccard coefficient similarity of the two sets of feature sub-vectors, the dissimilarity of the two sets of feature sub-vectors is calculated using the following formula: : Based on the dissimilarity Get the symmetric correlation matrix : The values ​​of the diagonal elements are 0.

[0038] It should be noted that the system constructs a symmetric correlation matrix by calculating the Jaccard coefficient similarity of any two sets of feature subvectors. The Jaccard coefficient is an indicator to measure the similarity of sets, which is calculated by comparing the size of the intersection and union of two sets. Here, the feature subvector refers to a subset of a specific dimension extracted from the original multidimensional feature vector. The symmetric correlation matrix is ​​a square matrix in which each element represents the similarity or dissimilarity between two feature subvectors.

[0039] Specifically, the system first determines the dimension sequence of any two sets of feature subvectors, and then calculates the number of dimensions contained in the two sequences (i.e., the size of the intersection) and the number of unique dimensions of each. This information is used to calculate the Jaccard coefficient, which is the number of common dimensions divided by the sum of the number of common dimensions plus the number of unique dimensions.

[0040] More specifically, the system uses the Jaccard coefficient to calculate the dissimilarity between any two sets of feature subvectors, that is, 1 minus the Jaccard coefficient. This dissimilarity is used to construct a symmetric correlation matrix, where each element of the matrix represents the dissimilarity between the corresponding feature subvector pair.

[0041] Preferably, when constructing the symmetric correlation matrix, the system can normalize the calculation results of the Jaccard coefficient to ensure that the values ​​of similarity and dissimilarity are between 0 and 1. In addition, the system can also use other similarity or dissimilarity calculation methods, such as cosine similarity, Euclidean distance, etc., to provide a more comprehensive analysis.

[0042] Furthermore, in actual operation, the system may use a matrix operation library in a programming language to efficiently calculate and store a symmetric correlation matrix, or adjust the calculation formula of the Jaccard coefficient according to a specific application scenario to adapt to different data characteristics.

[0043] In some embodiments, calculating the weights of the multiple groups of correlation results according to the symmetric correlation matrix includes: Using the following formula, multiple first weights are obtained according to the ratio of the sum of the element values ​​of each row in the matrix to the sum of all element values: The obtained multiple first weights are normalized and used as weights of multiple groups of correlation results.

[0044] It should be noted that the system calculates multiple first weights using the ratio of the sum of the values ​​of each row of the symmetric correlation matrix to the sum of all the values ​​of the elements. Here, the first weight refers to an indicator used to measure the importance of each set of feature sub-vectors in the overall data set in the correlation analysis. This ratio can help the system determine which feature sub-vectors have a greater impact on the final data matching results.

[0045] Specifically, the system first calculates the sum of the element values ​​of each row in the symmetric correlation matrix, and then calculates the sum of all element values ​​of the entire matrix. Next, the system divides the sum of the element values ​​of each row by the sum of all element values ​​to obtain the first weight of each feature subvector. During this process, the system may use matrix operation functions in the programming language to automate the calculation to ensure the accuracy and efficiency of the calculation. For example, if the symmetric correlation matrix is ​​a 5x5 matrix, the system will calculate the sum of each row and then divide it by 25 (5x5) to obtain five first weights.

[0046] Preferably, after obtaining multiple first weights, the system will perform normalization processing to ensure that the sum of all weights is 1. Normalization processing can be achieved by dividing each first weight by the sum of all first weights. This can make the weights more balanced in subsequent matching degree calculations and avoid some feature sub-vectors from excessively affecting the final result due to excessive weights.

[0047] Furthermore, the system can also consider introducing other factors, such as the stability or historical performance of the feature sub-vector, to adjust the calculation of the weights to improve the accuracy of the matching calculation. In actual operation, the system may use machine learning algorithms to dynamically adjust the weights to adapt to the changing data characteristics.

[0048] In some embodiments, the railway data matching count judgment condition is expressed as: in, is the data in the i-th group of correlation results , when the data If it is matched to the kth category, the count is 1, otherwise the count is 0.

[0049] It should be noted that the system calculates the data matching results by defining the railway data matching count judgment condition. Here, the railway data matching count judgment condition refers to a set of rules or standards used by the system to determine whether the data is successfully matched to a specific category. This condition involves a quantitative evaluation of the data matching results, usually based on the similarity or correlation between the data point and a specific category.

[0050] Specifically, the system assigns a match count to the data point in the i-th group of correlation results, where if the data point is successfully matched to the k-th category, the count is 1, otherwise it is 0. This counting process can be achieved by comparing the similarity between the feature vector of the data point and the reference feature vector of each category. For example, the system may use a distance metric (such as Euclidean distance) or a similarity metric (such as cosine similarity) to evaluate the similarity between the data point and each category, and set a threshold to decide whether to match the data point to a specific category.

[0051] Preferably, the system can adopt more complex match counting judgment conditions, such as considering the similarity between data points and multiple categories, and using a voting mechanism to determine the final matching result. In addition, the system can also introduce confidence scoring to assign a confidence score to each matching result, which can be obtained based on the correlation results of multiple feature sub-vectors.

[0052] Furthermore, in actual operation, the system may use a machine learning model to predict the category of a data point and adjust the matching count judgment condition based on the output of the model to improve the accuracy and robustness of the matching.

[0053] In some embodiments, the weighted matching degree of each category of each data in the multiple groups of correlation results is calculated according to the weight and the railway data matching count judgment condition, and the calculation formula is: in, For data The weighted matching degree of matching it to the kth category in the R group correlation results.

[0054] It should be noted that the system calculates the weighted matching degree of each category of each data through the weight and railway data matching count judgment conditions. Here, weighted matching degree refers to the data matching metric that takes into account the weight of the correlation result. It combines the similarity between the data point and each category and the weight of the correlation result. This metric can help the system more accurately determine which category the data point should belong to.

[0055] Specifically, the system calculates the weighted match based on the match of each data point in the R group of correlation results and the corresponding weight. This calculation can be achieved by multiplying the match count of each data point by the corresponding weight. For example, if a data point has a match count of 1 in the correlation result of the kth category, and the weight of the result is 0.7, then the weighted match of the data point in the kth category is 0.7. This calculation process can generate a weighted match value for each data point and each category.

[0056] Preferably, the system can further refine the calculation method of the weighted matching degree, for example, a nonlinear transformation can be introduced to process the product of the weight and the matching count to better reflect the actual matching degree between the data point and the category.

[0057] Furthermore, the system can also consider the distance between the feature subvector of the data point and the category center, and incorporate this distance as another factor into the calculation of the weighted matching degree. In actual operation, the system may use a comprehensive scoring mechanism that combines the correlation results and weights of multiple feature subvectors to determine the final category of each data point.

[0058] In some embodiments, in the data preprocessing, the number of nodes in the input layer is the number of dimensions of the railway data after preprocessing, the number of nodes in the output layer is the number of categories to which the railway data belongs, and the number of nodes in the last layer of the middle layer of the network is less than the number of dimensions of the input and greater than the number of categories of the output.

[0059] It should be noted that in data preprocessing, the system sets the number of nodes in the input layer to be the number of dimensions of the railway data after preprocessing, the number of nodes in the output layer to be the number of categories to which the railway data belongs, and the number of nodes in the last layer of the middle layer of the network is less than the number of dimensions of the input and greater than the number of categories of the output. Here, the number of nodes refers to the number of units or features contained in each layer in the data processing or machine learning model. The input layer, output layer, and middle layer are terms in neural networks, representing the levels at which the model receives input data, generates output results, and processes data.

[0060] Specifically, when the system preprocesses the railway data, it sets the structure of the neural network according to the characteristics of the data and the number of categories. For example, if the preprocessed railway data has 100 dimensions, and the railway data can be classified into 10 different categories, then the input layer will have 100 nodes and the output layer will have 10 nodes. The number of nodes in the middle layer is set according to actual needs, but must be less than 100 and greater than 10. Such a setting helps the model learn complex patterns in the data while avoiding overfitting. The specific parameter settings may depend on the specific characteristics of the data and the effect of model training.

[0061] Preferably, the system can use cross-validation and other techniques to determine the optimal network structure, including the number of nodes in the middle layer. In addition, the system can also consider using regularization techniques, such as L1 or L2 regularization, to prevent overfitting and improve the generalization ability of the model.

[0062] Furthermore, in actual operation, the system may optimize the model performance by adjusting the number of nodes in the middle layer and trying different activation functions, such as using the ReLU activation function to speed up training and reduce the problem of gradient disappearance. You can also consider using an optimization algorithm with adaptive learning rate, such as Adam, to improve training efficiency and model convergence speed.

[0063] In some embodiments, the preprocessing comprises: extracting key information fragments from the received railway data; Using principal component analysis PCA to reduce the dimension of the key information fragments; The preprocessed railway data is obtained by normalizing the reduced-dimensional railway data.

[0064] It should be noted that in the preprocessing step, the system extracts key information fragments from the received railway data and uses principal component analysis (PCA) to reduce the dimensionality of these key information fragments. Here, key information fragments refer to the data parts that are screened out from the original railway data and are of great value for subsequent analysis. Principal component analysis (PCA) is a statistical method used to transform potentially correlated variables into a set of linearly uncorrelated variables through orthogonal transformation. This process is usually used to reduce the dimensionality of data while retaining as much information as possible from the original data.

[0065] Specifically, the system first analyzes the received railway data to identify and extract data fragments containing key information. These fragments may include train operating parameters, maintenance logs, fault records, etc. Then, the system applies the PCA algorithm to reduce the dimensionality of these key information fragments to reduce the complexity of the data and extract the principal components that best represent the data characteristics. For example, if the original data set contains hundreds of features, PCA can help the system identify the most important features that can explain most of the variability in the data.

[0066] Preferably, after PCA dimensionality reduction, the system will also normalize the data to ensure that the comparison between different feature values ​​is fair. Normalization may include scaling the data to the range of [0,1] or [-1,1], or standardizing it to have zero mean and unit variance. Such processing helps to improve the accuracy of subsequent analysis and make model training more stable.

[0067] Furthermore, in actual operation, the system may combine other dimensionality reduction techniques, such as linear discriminant analysis (LDA) or t-distributed stochastic neighbor embedding (t-SNE), to further improve the separability and visualization of the data. In addition, the system can also adjust the parameters in the PCA algorithm, such as the number of principal components retained, according to the characteristics of the data and the analysis objectives, to achieve the best dimensionality reduction effect.

[0068] The above-mentioned embodiments of the present invention have the following beneficial effects: The railway data asset management system described in the present invention can significantly improve the matching accuracy and management efficiency of railway operation data and maintenance records by comprehensively using technologies such as data preprocessing, multi-dimensional feature vector generation, correlation analysis and weighted matching degree calculation. The system first receives railway operation data and maintenance records, performs data preprocessing to extract key information, and then generates multiple multi-dimensional feature vectors. By sampling these feature vectors with multiple randomly generated dimensions, multiple groups of feature sub-vectors are obtained, and correlation analysis is performed, which can deeply explore the complex correlations between data and provide a more accurate basis for railway data matching.

[0069] Furthermore, the system obtains a symmetric correlation matrix based on the attribute value differences between any two sets of feature sub-vectors in the multiple sets of feature sub-vectors, and calculates the weights of the multiple sets of correlation results. Combined with the railway data matching count judgment condition, the weighted matching degree of each category of each data in the multiple sets of correlation results can be calculated to determine the optimal matching category of the data. This method can not only optimize the data matching process, improve the accuracy and reliability of matching, but also enhance the intelligent level of railway data asset management, and provide more scientific and efficient data support for railway operations.

[0070] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0071] The above descriptions are only some preferred embodiments of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. A railway data asset management system, characterized in that: The steps include: receiving railway operation data and maintenance records, performing data preprocessing to extract key information, and using the preprocessed data to generate multiple multi-dimensional feature vectors; Sampling the plurality of multidimensional feature vectors according to the number of dimensions randomly generated multiple times to obtain a plurality of groups of feature sub-vectors, and performing correlation analysis on the plurality of groups of feature sub-vectors respectively to obtain a plurality of groups of correlation results; According to the difference in attribute values ​​between any two groups of feature sub-vectors in the multiple groups of feature sub-vectors, a symmetric correlation matrix is ​​obtained. According to the symmetric correlation matrix, the weights of the multiple groups of correlation results are calculated. According to the weights and the railway data matching count judgment condition, the weighted matching degree of each category of each data in the multiple groups of correlation results is calculated, and the category corresponding to the maximum weighted matching degree is taken as the optimal matching category of the data.

2. The railway data asset management system according to claim 1, characterized in that: The following steps are involved: Performing statistical analysis on the dimensions of the correlation results to determine the distribution range of key dimensions; Based on the importance and distribution range of the key dimensions, 70% to 80% of the key dimensions are selected as the number of dimensions, which is used as the basis for sampling; On the basis of the benchmark, multiple values ​​are selected by random sampling method, the multiple values ​​are used as the number of dimensions, the number of values ​​is used as the sampling times, and the sampling times is no less than 3 times.

3. The railway data asset management system according to claim 2, characterized in that: The sampling of the plurality of multidimensional feature vectors according to the number of dimensions randomly generated for a plurality of times to obtain a plurality of groups of feature sub-vectors includes: According to the dimension number ci obtained by the i-th sampling, a non-repeating number ci is randomly generated within the dimension range of the multidimensional feature vector, and after being arranged from small to large, the eigenvalues ​​of the corresponding dimensional sequence position are extracted for each multidimensional feature vector as a dimensional sequence to obtain the i-th group of feature sub-vectors; The feature sub-vectors obtained by each sampling constitute multiple groups of feature sub-vectors.

4. The railway data asset management system according to any one of claims 1 to 3, characterized in that: The correlation analysis adopts the improved Pearson correlation coefficient algorithm to calculate two initial correlation centers according to the attribute value of each dimension in the feature sub-vector. The formula is: in, and are the attribute values ​​of the rth dimension in the feature sub-vector, and They are and The average value of is the total number of dimensions, r is the correlation coefficient between X and Y, X is the attribute value of the Xth group of feature sub-vectors, and Y is the attribute value of the Yth group of feature sub-vectors.

5. The railway data asset management system according to claim 3, characterized in that: The method of obtaining a symmetric correlation matrix according to the attribute value difference between any two groups of feature sub-vectors in the plurality of groups of feature sub-vectors comprises: The Jaccard coefficient similarity of any two sets of feature sub-vectors is calculated by the following formula: in, and are the dimension sequences of any two groups of feature sub-vectors, for and The number of dimensions contained in for Contains, and The number of dimensions not included, for Contains, and The number of dimensions not included; Based on the Jaccard coefficient similarity of the two sets of feature sub-vectors, the dissimilarity of the two sets of feature sub-vectors is calculated using the following formula: : Based on the dissimilarity Get the symmetric correlation matrix : The values ​​of the diagonal elements are 0.

6. The railway data asset management system according to claim 5, characterized in that: The step of calculating the weights of the plurality of groups of correlation results according to the symmetric correlation matrix comprises: Using the following formula, multiple first weights are obtained according to the ratio of the sum of the element values ​​of each row in the matrix to the sum of all element values: The obtained multiple first weights are normalized and used as weights of multiple groups of correlation results.

7. The railway data asset management system according to claim 1 or 6, characterized in that: The railway data matching count judgment condition is expressed as: in, is the data in the i-th group of correlation results , when the data If it is matched to the kth category, the count is 1, otherwise the count is 0.

8. The railway data asset management system according to claim 7, characterized in that: According to the weight and railway data matching count judgment condition, the weighted matching degree of each category of each data in the multiple groups of correlation results is calculated, and the calculation formula is: in, For data The weighted matching degree of matching it to the kth category in the R group correlation results.

9. The railway data asset management system according to claim 8, characterized in that: In the data preprocessing, the number of nodes in the input layer is the number of dimensions of the railway data after preprocessing, the number of nodes in the output layer is the number of categories to which the railway data belongs, and the number of nodes in the last layer of the middle layer of the network is less than the number of dimensions of the input and greater than the number of categories of the output.

10. The railway data asset management system according to claim 9, characterized in that: The pre-processing comprises: extracting key information fragments from the received railway data; Using principal component analysis PCA to reduce the dimension of the key information fragments; The preprocessed railway data is obtained by normalizing the reduced-dimensional railway data.

Images