Rail transit data analysis method and system based on adaptive feature engineering

Abstract

The present invention provides a rail transit data analysis method and system based on adaptive feature engineering, the method comprising: obtaining rail transit data to be analyzed; using a preset adaptive feature engineering algorithm to perform deep feature extraction on the rail transit data to be analyzed to obtain deep feature data; using a preset quality control algorithm to perform abnormality detection and / or data repair on the deep feature data to obtain quality-controlled data; using a preset analysis and processing algorithm to perform parallel processing and analysis on the quality-controlled data to obtain rail transit data analysis results; wherein the adaptive feature engineering algorithm, the quality control algorithm, and the analysis and processing algorithm are dynamically optimized by deep transfer learning. The present invention can fully tap the deep information contained in rail transit data, significantly improve the processing efficiency and analysis accuracy of rail transit data, and provide reliable technical support for the digital transformation of the industry.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and provides a rail transit data analysis method and system based on adaptive feature engineering. Background Art

[0002] With the rapid development of the rail transit industry, rail transit data has shown an explosive growth trend, with characteristics such as multi-source heterogeneity, time series correlation, and unstable quality. Traditional data processing methods can no longer meet the needs of the digital transformation of the rail transit industry. The data processing systems currently used in the market have many limitations and show obvious performance bottlenecks when facing TB-level rail transit data. Although they can complete basic data processing tasks, they cannot fully explore the deep information contained in rail transit data. Because its universal design cannot adapt well to the particularity of rail transit industry data, it leads to inefficient utilization of system resources. In addition, due to the lack of an effective quality control mechanism, the reliability of the output analysis results of the existing processing flow is difficult to guarantee. Summary of the invention

[0003] The present invention provides a rail transit data analysis method and system based on adaptive feature engineering, which is used to solve the defects of the prior art in processing rail transit data, such as inaccurate feature extraction, low processing efficiency, unreliable quality assessment, etc., and is difficult to adapt to the needs of digital transformation of the rail transit industry. The present invention can fully mine the deep information contained in rail transit data, significantly improve the processing efficiency and analysis accuracy of rail transit data, and provide reliable technical support for the digital transformation of the industry.

[0004] The present invention provides a rail transit data analysis method based on adaptive feature engineering, comprising: acquiring rail transit data to be analyzed; using a preset adaptive feature engineering algorithm to perform deep feature extraction on the rail transit data to be analyzed to obtain deep feature data; the adaptive feature engineering algorithm comprises a combination of one or more of an adaptive interpolation algorithm, an adaptive decomposition algorithm, an adaptive frequency domain feature extraction algorithm and an adaptive feature fusion algorithm; using a preset quality control algorithm to perform anomaly detection and / or data repair on the deep feature data to obtain quality-controlled data; the quality control algorithm comprises an anomaly detection algorithm and / or a data repair algorithm; using a preset analysis and processing algorithm to perform parallel processing and analysis on the quality-controlled data to obtain rail transit data analysis results; the analysis and processing algorithm comprises a combination of one or more of a dynamic load balancing algorithm, a task similarity analysis algorithm and an elastic scaling algorithm; wherein the adaptive feature engineering algorithm, the quality control algorithm and the analysis and processing algorithm perform algorithm dynamic optimization by means of deep transfer learning.

[0005] According to a rail transit data analysis method based on adaptive feature engineering provided by the present invention, before the deep feature extraction of the rail transit data to be analyzed is performed using a preset adaptive feature engineering algorithm, and the deep feature data is obtained, the method also includes: performing input processing on the rail transit data to be analyzed using a preset data input algorithm; obtaining the rail transit data after input processing so as to perform deep feature extraction; the data input algorithm includes a data access algorithm and a data preprocessing algorithm.

[0006] According to a rail transit data analysis method based on adaptive feature engineering provided by the present invention, the method uses a preset adaptive feature engineering algorithm to perform deep feature extraction on the rail transit data to be analyzed to obtain deep feature data, including: using the adaptive interpolation algorithm to process missing values ​​in the rail transit data to be analyzed to obtain interpolated rail transit data; the adaptive interpolation algorithm constructs a state transfer equation based on the dynamic programming idea, and captures the trend and periodic characteristics of the data through polynomial regression and Fourier decomposition; using the adaptive decomposition algorithm to perform wavelet packet decomposition on the interpolated rail transit data to obtain decomposed rail transit data; the adaptive decomposition algorithm takes the weighted sum of reconstruction error and energy entropy as the optimization target; using the adaptive frequency domain feature extraction algorithm to extract frequency domain features of the decomposed rail transit data to obtain frequency domain features of the rail transit data; the adaptive frequency domain feature extraction algorithm adaptively allocates frequency band energy, and selects features through feature correlation and feature importance; using the adaptive feature fusion algorithm to fuse the frequency domain features of the rail transit data, and using the fused features as the deep feature data; the adaptive feature fusion algorithm captures the correlation between features based on the attention mechanism.

[0007] According to a rail transit data analysis method based on adaptive feature engineering provided by the present invention, the method adopts a preset quality control algorithm to perform anomaly detection and / or data repair on the deep feature data to obtain quality-controlled data, including: adopting the anomaly detection algorithm to perform anomaly detection on the deep feature data to obtain data after anomaly detection; the anomaly detection algorithm is a local sensitive hashing algorithm; adopting the data repair algorithm to perform data repair on the data after anomaly detection, and using the repaired data as the quality-controlled data; the data repair algorithm adopts a deep generative model to reconstruct the abnormal data; the deep generative model uses rail transit domain knowledge as a constraint.

[0008] According to a rail transit data analysis method based on adaptive feature engineering provided by the present invention, the preset analysis and processing algorithm is used to perform parallel processing and analysis on the data after quality control to obtain the rail transit data analysis result, including: using the dynamic load balancing algorithm to allocate computing tasks for the data after quality control to obtain a plurality of computing tasks; using the task similarity analysis algorithm to perform similarity analysis on a plurality of the computing tasks to obtain a similarity analysis result; and according to the similarity analysis result, using the elastic scaling algorithm to manage task computing resources to obtain the rail transit data analysis result.

[0009] According to a rail transit data analysis method based on adaptive feature engineering provided by the present invention, the loss function of the deep transfer learning is: in, L is the loss function of the deep transfer learning, L c is the classification loss function, L d To combat the loss function, is a trade-off parameter.

[0010] The present invention also provides a rail transit data analysis system based on adaptive feature engineering, comprising: an acquisition module, used to acquire rail transit data to be analyzed; a feature extraction module, used to perform deep feature extraction on the rail transit data to be analyzed using a preset adaptive feature engineering algorithm to obtain deep feature data; the adaptive feature engineering algorithm comprises a combination of one or more of an adaptive interpolation algorithm, an adaptive decomposition algorithm, an adaptive frequency domain feature extraction algorithm and an adaptive feature fusion algorithm; a quality control module, used to perform anomaly detection and / or data repair on the deep feature data using a preset quality control algorithm to obtain quality-controlled data; the quality control algorithm comprises an anomaly detection algorithm and / or a data repair algorithm; an analysis and processing module, used to perform parallel processing and analysis on the quality-controlled data using a preset analysis and processing algorithm to obtain rail transit data analysis results; the analysis and processing algorithm comprises a combination of one or more of a dynamic load balancing algorithm, a task similarity analysis algorithm and an elastic scaling algorithm; wherein the adaptive feature engineering algorithm, the quality control algorithm and the analysis and processing algorithm perform algorithm dynamic optimization by means of deep transfer learning.

[0011] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the rail transit data analysis method based on adaptive feature engineering as described in any one of the above-mentioned methods is implemented.

[0012] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for analyzing rail transit data based on adaptive feature engineering as described in any one of the above is implemented.

[0013] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned rail transit data analysis methods based on adaptive feature engineering.

[0014] The present invention provides a rail transit data analysis method and system based on adaptive feature engineering, the method comprising: obtaining rail transit data to be analyzed; using a preset adaptive feature engineering algorithm to perform deep feature extraction on the rail transit data to be analyzed to obtain deep feature data; using a preset quality control algorithm to perform anomaly detection and / or data repair on the deep feature data to obtain quality-controlled data; using a preset analysis and processing algorithm to perform parallel processing and analysis on the quality-controlled data to obtain rail transit data analysis results; wherein the adaptive feature engineering algorithm, the quality control algorithm, and the analysis and processing algorithm are dynamically optimized by deep transfer learning. The present invention can fully tap the deep information contained in rail transit data, significantly improve the processing efficiency and analysis accuracy of rail transit data, and provide reliable technical support for the digital transformation of the industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 It is a flow chart of the rail transit data analysis method based on adaptive feature engineering provided by the present invention.

[0017] Figure 2 It is a structural schematic diagram of the rail transit data analysis system based on adaptive feature engineering provided by the present invention.

[0018] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0019] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0020] Please refer to Figure 1 , Figure 1 A schematic flow chart of the rail transit data analysis method based on adaptive feature engineering provided by the present invention.

[0021] The present invention provides a rail transit data analysis method based on adaptive feature engineering, comprising: 101: Obtain rail transit data to be analyzed.

[0022] As a preferred embodiment, a preset adaptive feature engineering algorithm is used to perform deep feature extraction on the rail transit data to be analyzed. Before obtaining the deep feature data, it also includes: using a preset data input algorithm to perform input processing on the rail transit data to be analyzed; obtaining the rail transit data after input processing for deep feature extraction; the data input algorithm includes a data access algorithm and a data preprocessing algorithm.

[0023] In this embodiment, a preset data input algorithm is used to process data streams from different devices. The data access algorithm designs an extensible data access protocol that supports multiple communication methods such as TCP / IP (Transmission Control Protocol / Internet Protocol), serial port, and CAN (Controller Area Network) bus. In protocol parsing, a state machine method is used to identify the data frame format, and a checksum mechanism is used to ensure data integrity. Based on Zero-Copy technology, the data transmission efficiency is optimized, and a memory mapping method is used to achieve high-speed data caching. The data preprocessing algorithm is used to unify the data format, and through a dynamic data model, it can adapt to different types of data structures. The dynamic data model defines data specifications based on JSON Schema, and realizes dynamic mapping of data fields through a reflection mechanism. The NTP (Network Time Protocol) algorithm is used to process clock deviations in a distributed environment, achieving microsecond data alignment accuracy. A data flow control mechanism is introduced to dynamically adjust the data acquisition frequency according to the processing capacity to avoid system overload. The preset data input algorithm is used to input and process the rail transit data to be analyzed; the input processed rail transit data is obtained for deep feature extraction.

[0024] 102: Use a preset adaptive feature engineering algorithm to perform deep feature extraction on the rail transit data to be analyzed to obtain deep feature data; the adaptive feature engineering algorithm includes one or more combinations of an adaptive interpolation algorithm, an adaptive decomposition algorithm, an adaptive frequency domain feature extraction algorithm, and an adaptive feature fusion algorithm.

[0025] As a preferred embodiment, a preset adaptive feature engineering algorithm is used to perform deep feature extraction on the rail transit data to be analyzed to obtain deep feature data, including: using an adaptive interpolation algorithm to process missing values ​​in the rail transit data to be analyzed to obtain interpolated rail transit data; the adaptive interpolation algorithm constructs a state transfer equation based on the idea of ​​dynamic programming, and captures the trend and periodic characteristics of the data through polynomial regression and Fourier decomposition; an adaptive decomposition algorithm is used to perform wavelet packet decomposition on the interpolated rail transit data to obtain decomposed rail transit data; the adaptive decomposition algorithm takes the weighted sum of reconstruction error and energy entropy as the optimization target; an adaptive frequency domain feature extraction algorithm is used to extract frequency domain features of the decomposed rail transit data to obtain frequency domain features of the rail transit data; the adaptive frequency domain feature extraction algorithm adaptively allocates frequency band energy, and selects features through feature correlation and feature importance; an adaptive feature fusion algorithm is used to fuse the frequency domain features of the rail transit data, and the fused features are used as deep feature data; the adaptive feature fusion algorithm captures the correlation between features based on the attention mechanism.

[0026] In this embodiment, a preset adaptive feature engineering algorithm is used to perform deep feature extraction on the rail transit data to be analyzed to obtain deep feature data. Specifically, data cleaning is first performed on the rail transit data to be analyzed, and an adaptive interpolation algorithm is used to process missing values ​​to obtain interpolated rail transit data. The adaptive interpolation algorithm constructs a state transfer equation based on the idea of ​​dynamic programming, and captures the trend and periodic characteristics of the data through polynomial regression and Fourier decomposition. The adaptive interpolation algorithm introduces an adaptive weight mechanism, and the weight coefficient is dynamically adjusted according to the time distance and correlation. The calculation formula is: in, w ( i ) is the i The weight value of each sample point; t is the time point to be interpolated; t i For the i The time of the known sample points; λ is the time decay factor, which controls the influence of time distance; corr ( x ( t ), x ( t i )) is the correlation coefficient between the point to be interpolated and the known sample point; exp () is the natural exponential function.

[0027] Integrate multiple basic interpolators and use weighted voting to improve interpolation robustness.

[0028] In order to solve the problem that it is difficult to unify the parameters of feature extraction of different types of signals, an adaptive decomposition algorithm is used to decompose the interpolated rail transit data by wavelet packets to obtain the decomposed rail transit data. The accurate decomposition of different types of signals is achieved by constructing a dynamic basis function library. The basis function optimization algorithm takes the weighted sum of reconstruction error and energy entropy as the optimization target, and the optimization function is defined as: Among them, J(ψ) is the optimization objective function; ψ is the wavelet basis function; E(ψ) is the signal reconstruction error; H(ψ) is the energy entropy of the decomposition coefficient; α is the balance factor, and its value range is [0, 1].

[0029] The adaptive decomposition algorithm dynamically adjusts the wavelet basis function parameters in each decomposition layer, and the parameter update adopts the gradient descent method with momentum term. The adaptive decomposition depth control mechanism is implemented, and the optimal number of decomposition layers is determined based on the inter-layer energy ratio and information gain.

[0030] The adaptive frequency domain feature extraction algorithm is used to extract the frequency domain features of the decomposed rail transit data to obtain the frequency domain features of the rail transit data. The adaptive frequency domain feature extraction algorithm optimizes the resolution by dynamically adjusting the time-frequency window. The window size is based on the Hurst exponent to evaluate the local characteristics of the signal and realize the adaptive allocation of frequency band energy. In order to solve the limitation of feature engineering that relies heavily on expert experience, in feature selection, a multi-objective optimization strategy based on graph structure is used to express feature dependencies through feature correlation graphs. The PageRank algorithm used for feature importance evaluation considers both the inherent importance and structural importance of the feature. A dynamic update mechanism for feature importance is introduced to adjust feature weights in real time according to changes in data distribution.

[0031] In order to solve the technical problem of insufficient feature correlation analysis of multi-source heterogeneous data, an adaptive feature fusion algorithm is used to fuse the frequency domain features of rail transit data, and the fused features are used as deep feature data. The adaptive feature fusion algorithm is an adaptive fusion strategy based on the attention mechanism. The bilinear pooling operation used in the attention weight calculation effectively captures the high-order correlation between features. In the feature semantic embedding, the mapping of features to semantic space is achieved through pre-training feature encoders. The fusion process uses a gated recurrent unit (GRU) to model feature sequence dependency and extract temporal correlation information. A feature conflict detection mechanism is also implemented to identify and resolve conflicts between features through the information entropy criterion.

[0032] 103: Use a preset quality control algorithm to perform anomaly detection and / or data repair on the deep feature data to obtain quality controlled data; the quality control algorithm includes an anomaly detection algorithm and / or a data repair algorithm.

[0033] As a preferred embodiment, a preset quality control algorithm is used to perform anomaly detection and / or data repair on deep feature data to obtain quality controlled data, including: using an anomaly detection algorithm to perform anomaly detection on deep feature data to obtain data after anomaly detection; the anomaly detection algorithm is a local sensitive hashing algorithm; using a data repair algorithm to repair the data after anomaly detection, and using the repaired data as quality controlled data; the data repair algorithm uses a deep generative model to reconstruct the abnormal data; the deep generative model uses rail transit domain knowledge as a constraint.

[0034] In this embodiment, a preset quality control algorithm is used to perform anomaly detection and / or data repair on the deep feature data to obtain quality controlled data.

[0035] In order to solve the technical problems of high computational overhead and poor real-time performance of traditional methods, an anomaly detection algorithm is used to detect anomalies in deep feature data to obtain data after anomaly detection. The anomaly detection algorithm is an efficient detection algorithm based on locality-sensitive hashing (LSH) technology. The hash function family adopts a projection plus quantization strategy to ensure the mapping similarity of spatially close points. A dynamic hash table update mechanism is introduced to adaptively adjust parameters by monitoring the hash conflict rate. The detection process integrates the local outlier factor (LOF) detector, isolation forest detector and deep autoencoder detector, and uses an online learning method to dynamically adjust the detector weights.

[0036] The data repair algorithm is used to repair the data after anomaly detection, and the repaired data is used as the data after quality control. The data repair algorithm uses a deep generative model to reconstruct abnormal data. The discriminator network of the deep generative model introduces domain knowledge constraints to ensure that the generated data meets the laws of physics. The data enhancement function based on the conditional generative adversarial network (CGAN) is implemented, and the data distribution is optimized through difficult example mining and active learning strategies. In order to solve the problem of poor scalability of existing evaluation methods, in quality evaluation, based on the rule engine design, the decision tree is used to extract evaluation rules from historical records. The rule optimization algorithm eliminates redundant rules and improves evaluation efficiency.

[0037] 104: Use a preset analysis and processing algorithm to perform parallel processing and analysis on the quality-controlled data to obtain rail transit data analysis results; the analysis and processing algorithm includes a combination of one or more of a dynamic load balancing algorithm, a task similarity analysis algorithm, and an elastic scaling algorithm; wherein the adaptive feature engineering algorithm, the quality control algorithm, and the analysis and processing algorithm are dynamically optimized by deep transfer learning.

[0038] As a preferred embodiment, a preset analysis and processing algorithm is used to perform parallel processing and analysis on the data after quality control to obtain rail transit data analysis results, including: using a dynamic load balancing algorithm to allocate computing tasks for the data after quality control to obtain a number of computing tasks; using a task similarity analysis algorithm to perform similarity analysis on a number of computing tasks to obtain similarity analysis results; based on the similarity analysis results, an elastic scaling algorithm is used to manage task computing resources to obtain rail transit data analysis results.

[0039] In this embodiment, a preset analysis and processing algorithm is used to perform parallel processing and analysis on the quality-controlled data to obtain rail transit data analysis results.

[0040] In order to solve the problem of low efficiency in large-scale data parallel processing, a dynamic load balancing algorithm is used to allocate the computing tasks of the quality-controlled data, and several computing tasks are obtained, breaking through the technical bottleneck of uneven load and low resource utilization of traditional task allocation methods. The dynamic load balancing algorithm optimizes task allocation through the resource utilization model. In time prediction, the LSTM (Long Short-Term Memory) network structure is used to predict the task execution time. In order to solve the problem of high computational overhead in the task scheduling optimization process, the task similarity analysis algorithm is used to perform similarity analysis on several computing tasks, and the similarity analysis results are obtained to guide resource allocation. According to the similarity analysis results, the elastic scaling algorithm is used to manage the task computing resources to obtain the rail transit data analysis results. The elastic scaling algorithm predicts resource requirements by analyzing the utilization curve. It realizes the capture of periodic laws based on the sliding time window, optimizes the resource scheduling strategy, and has high reliability.

[0041] As a preferred embodiment, the loss function of deep transfer learning is: in, L is the loss function of deep transfer learning, L c is the classification loss function, L d To combat the loss function, is a trade-off parameter.

[0042] In order to solve the problem that the performance of the prior art drops sharply when processing scenarios with large domain offsets, in this embodiment, deep transfer learning runs through all processing links, and cross-condition knowledge transfer is achieved through deep transfer learning. Adversarial training is used to extract domain-independent features in domain adaptation. The loss function of deep transfer learning is used to measure the overall error of the model. It consists of multiple parts, which comprehensively consider the performance of the model on different targets. The classification loss function is the error of the model when performing the main task (such as classification, regression, etc.), usually measured by cross entropy loss or mean squared error (MSE). The adversarial loss function is used to extract domain-independent features through adversarial training, so that the model can adapt between different domains. It is often used in knowledge transfer to ensure that similar features are extracted between the source domain and the target domain, and reduce domain deviation. The trade-off parameter controls the balance between the classification loss function and the adversarial loss function. By adjusting the value of the trade-off parameter, the importance of the classification loss and the domain adaptation loss in the overall loss can be adjusted to ensure that the model can maintain good classification performance and effectively adapt to different domains.

[0043] In order to solve the problem of migration failure caused by large differences in feature distribution, an attention-enhanced feature fusion mechanism is introduced to capture feature correlation through bilinear pooling. Knowledge migration adopts a multi-level strategy. Shallow features are aligned through the MMD (Maximum Mean Discrepancy) criterion, and deep knowledge is migrated through knowledge distillation adjusted by temperature parameters. The migration effect is monitored through JS (Jensen-Shannon Divergence) divergence to achieve dynamic optimization of the migration strategy.

[0044] The rail transit data analysis system based on adaptive feature engineering provided by the present invention is described below. The rail transit data analysis system based on adaptive feature engineering described below and the rail transit data analysis method based on adaptive feature engineering described above can be referenced to each other.

[0045] Please refer to Figure 2 , Figure 2 A schematic diagram of the structure of the rail transit data analysis system based on adaptive feature engineering provided by the present invention.

[0046] The present invention also provides a rail transit data analysis system based on adaptive feature engineering, comprising: an acquisition module 201, used to acquire rail transit data to be analyzed; a feature extraction module 202, used to perform deep feature extraction on the rail transit data to be analyzed using a preset adaptive feature engineering algorithm to obtain deep feature data; the adaptive feature engineering algorithm comprises a combination of one or more of an adaptive interpolation algorithm, an adaptive decomposition algorithm, an adaptive frequency domain feature extraction algorithm and an adaptive feature fusion algorithm; a quality control module 203, used to perform anomaly detection and / or data repair on the deep feature data using a preset quality control algorithm to obtain quality-controlled data; the quality control algorithm comprises an anomaly detection algorithm and / or a data repair algorithm; an analysis and processing module 204, used to perform parallel processing and analysis on the quality-controlled data using a preset analysis and processing algorithm to obtain rail transit data analysis results; the analysis and processing algorithm comprises a combination of one or more of a dynamic load balancing algorithm, a task similarity analysis algorithm and an elastic scaling algorithm; wherein the adaptive feature engineering algorithm, the quality control algorithm and the analysis and processing algorithm perform dynamic algorithm optimization by means of deep transfer learning.

[0047] The present invention provides an intelligent, adaptive and efficient rail transit data analysis system. By building a complete technical solution, the system comprehensively solves the technical problems of key links such as feature engineering, quality control, distributed computing and knowledge transfer, adopts a layered adaptive architecture to achieve end-to-end data processing, and continuously optimizes the processing strategy through the knowledge transfer learning model, which significantly improves the processing efficiency and analysis accuracy of rail transit data, and provides reliable technical support for the digital transformation of the industry. The system realizes inter-layer data transmission through distributed message queues, uses distributed cache to accelerate data access, and stores processing results based on distributed file systems.

[0048] The end-to-end processing flow of the system is illustrated by taking the vehicle braking system test as an example. The acquisition module receives multi-source data streams from brake pressure sensors, acceleration sensors, and temperature sensors, and realizes data alignment through protocol analysis and time synchronization. The feature extraction module applies wavelet packet decomposition to the brake pressure curve to extract time-frequency features, analyzes the braking performance in combination with the acceleration signal, and evaluates the thermal attenuation characteristics of the brake through temperature data. The quality control module identifies and repairs abnormal data caused by sensor failure to ensure the reliability of the analysis results. The analysis and processing module trains the brake performance evaluation model based on historical data, and promotes the evaluation experience to new vehicle platforms through knowledge transfer.

[0049] The system was put into use during the trial production of a certain type of subway vehicle and successfully analyzed more than 10,000 hours of brake system data. Through in-depth analysis of features such as brake pressure, braking distance, and wheel slip, the system promptly discovered several brake system design defects, providing an important basis for optimizing the design solution. In the subsequent development of new models, the system effectively shortened the cycle through knowledge transfer and reduced the brake system debugging time by one third.

[0050] Figure 3 An example of a structural diagram of an electronic device is shown in FIG. Figure 3 As shown, the electronic device may include: a processor (processor) 301, a communication interface (Communications Interface) 302, a memory (memory) 303 and a communication bus 304, wherein the processor 301, the communication interface 302, and the memory 303 communicate with each other through the communication bus 304. The processor 301 can call the logic instructions in the memory 303 to execute the rail transit data analysis method based on adaptive feature engineering, which includes: obtaining the rail transit data to be analyzed; using a preset adaptive feature engineering algorithm to perform deep feature extraction on the rail transit data to be analyzed to obtain deep feature data; the adaptive feature engineering algorithm includes a combination of one or more of an adaptive interpolation algorithm, an adaptive decomposition algorithm, an adaptive frequency domain feature extraction algorithm and an adaptive feature fusion algorithm; using a preset quality control algorithm to perform anomaly detection and / or data repair on the deep feature data to obtain quality-controlled data; the quality control algorithm includes an anomaly detection algorithm and / or a data repair algorithm; using a preset analysis and processing algorithm to perform parallel processing and analysis on the quality-controlled data to obtain rail transit data analysis results; the analysis and processing algorithm includes a combination of one or more of a dynamic load balancing algorithm, a task similarity analysis algorithm and an elastic scaling algorithm; wherein the adaptive feature engineering algorithm, the quality control algorithm and the analysis and processing algorithm perform dynamic algorithm optimization through deep transfer learning.

[0051] In addition, the logic instructions in the above-mentioned memory 303 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the relevant technology or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0052] The embodiment of the present invention discloses a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the rail transit data analysis method based on adaptive feature engineering provided by the above-mentioned method embodiments, and the method includes: obtaining rail transit data to be analyzed; using a preset adaptive feature engineering algorithm to perform deep feature extraction on the rail transit data to be analyzed to obtain deep feature data; the adaptive feature engineering algorithm includes a combination of one or more of an adaptive interpolation algorithm, an adaptive decomposition algorithm, an adaptive frequency domain feature extraction algorithm and an adaptive feature fusion algorithm; using a preset quality control algorithm to perform anomaly detection and / or data repair on the deep feature data to obtain quality-controlled data; the quality control algorithm includes an anomaly detection algorithm and / or a data repair algorithm; using a preset analysis and processing algorithm to perform parallel processing and analysis on the quality-controlled data to obtain rail transit data analysis results; the analysis and processing algorithm includes a combination of one or more of a dynamic load balancing algorithm, a task similarity analysis algorithm and an elastic scaling algorithm; wherein the adaptive feature engineering algorithm, the quality control algorithm and the analysis and processing algorithm perform algorithm dynamic optimization by means of deep transfer learning.

[0053] On the other hand, an embodiment of the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the rail transit data analysis method based on adaptive feature engineering provided by the above-mentioned embodiments, the method comprising: obtaining rail transit data to be analyzed; using a preset adaptive feature engineering algorithm to perform deep feature extraction on the rail transit data to be analyzed to obtain deep feature data; the adaptive feature engineering algorithm includes a combination of one or more of an adaptive interpolation algorithm, an adaptive decomposition algorithm, an adaptive frequency domain feature extraction algorithm and an adaptive feature fusion algorithm; using a preset quality control algorithm to perform anomaly detection and / or data repair on the deep feature data to obtain quality-controlled data; the quality control algorithm includes an anomaly detection algorithm and / or a data repair algorithm; using a preset analysis and processing algorithm to perform parallel processing and analysis on the quality-controlled data to obtain rail transit data analysis results; the analysis and processing algorithm includes a combination of one or more of a dynamic load balancing algorithm, a task similarity analysis algorithm and an elastic scaling algorithm; wherein the adaptive feature engineering algorithm, the quality control algorithm and the analysis and processing algorithm perform dynamic algorithm optimization by means of deep transfer learning.

[0054] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0055] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these 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.

Claims

1. A rail transit data analysis method based on adaptive feature engineering, characterized in that: include: Obtain rail transit data to be analyzed; A preset adaptive feature engineering algorithm is used to perform deep feature extraction on the rail transit data to be analyzed to obtain deep feature data; the adaptive feature engineering algorithm includes a combination of one or more of an adaptive interpolation algorithm, an adaptive decomposition algorithm, an adaptive frequency domain feature extraction algorithm and an adaptive feature fusion algorithm; Using a preset quality control algorithm to perform anomaly detection and / or data repair on the depth feature data to obtain quality controlled data; the quality control algorithm includes an anomaly detection algorithm and / or a data repair algorithm; A preset analysis and processing algorithm is used to perform parallel processing and analysis on the quality-controlled data to obtain rail transit data analysis results; the analysis and processing algorithm includes a combination of one or more of a dynamic load balancing algorithm, a task similarity analysis algorithm, and an elastic scaling algorithm; Among them, the adaptive feature engineering algorithm, the quality control algorithm and the analysis and processing algorithm are dynamically optimized by means of deep transfer learning.

2. The rail transit data analysis method based on adaptive feature engineering according to claim 1 is characterized in that: Before the method of extracting deep features from the rail transit data to be analyzed by using a preset adaptive feature engineering algorithm to obtain deep feature data, the method further includes: The rail transit data to be analyzed is input and processed using a preset data input algorithm; the rail transit data after input processing is obtained so as to perform deep feature extraction; the data input algorithm includes a data access algorithm and a data preprocessing algorithm.

3. The rail transit data analysis method based on adaptive feature engineering according to claim 1 is characterized in that: The method of using a preset adaptive feature engineering algorithm to extract deep features from the rail transit data to be analyzed to obtain deep feature data includes: The adaptive interpolation algorithm is used to process the missing values ​​in the rail transit data to be analyzed to obtain the interpolated rail transit data; the adaptive interpolation algorithm constructs a state transfer equation based on the dynamic programming idea, and captures the trend and periodic characteristics of the data through polynomial regression and Fourier decomposition; The adaptive decomposition algorithm is used to perform wavelet packet decomposition on the interpolated rail transit data to obtain decomposed rail transit data; the adaptive decomposition algorithm takes the weighted sum of reconstruction error and energy entropy as an optimization target; The adaptive frequency domain feature extraction algorithm is used to extract frequency domain features from the decomposed rail transit data to obtain frequency domain features of the rail transit data; the adaptive frequency domain feature extraction algorithm adaptively allocates frequency band energy and performs feature selection based on feature correlation and feature importance; The adaptive feature fusion algorithm is used to fuse the frequency domain features of the rail transit data, and the fused features are used as the deep feature data; the adaptive feature fusion algorithm captures the correlation between features based on the attention mechanism.

4. The rail transit data analysis method based on adaptive feature engineering according to claim 1 is characterized in that: The using of a preset quality control algorithm to perform anomaly detection and / or data repair on the depth feature data to obtain quality controlled data includes: The anomaly detection algorithm is used to perform anomaly detection on the deep feature data to obtain data after anomaly detection; the anomaly detection algorithm is a local sensitive hashing algorithm; The data repair algorithm is used to repair the data after the anomaly detection, and the repaired data is used as the data after quality control; the data repair algorithm uses a deep generation model to reconstruct the abnormal data; the deep generation model uses rail transit field knowledge as a constraint condition.

5. The rail transit data analysis method based on adaptive feature engineering according to claim 1 is characterized in that: The method of using a preset analysis and processing algorithm to perform parallel processing and analysis on the quality-controlled data to obtain rail transit data analysis results includes: The dynamic load balancing algorithm is used to allocate computing tasks for the quality-controlled data to obtain a plurality of computing tasks; Using the task similarity analysis algorithm to perform similarity analysis on a plurality of the computing tasks to obtain similarity analysis results; According to the similarity analysis result, the elastic scaling algorithm is used to manage task computing resources to obtain the rail transit data analysis result.

6. The rail transit data analysis method based on adaptive feature engineering according to any one of claims 1 to 5, characterized in that: The loss function of the deep transfer learning is: in, L is the loss function of the deep transfer learning, L c is the classification loss function, L d To combat the loss function, is a trade-off parameter.

7. A rail transit data analysis system based on adaptive feature engineering, characterized in that: include: An acquisition module, used to acquire rail transit data to be analyzed; A feature extraction module, used to perform deep feature extraction on the rail transit data to be analyzed using a preset adaptive feature engineering algorithm to obtain deep feature data; the adaptive feature engineering algorithm includes a combination of one or more of an adaptive interpolation algorithm, an adaptive decomposition algorithm, an adaptive frequency domain feature extraction algorithm, and an adaptive feature fusion algorithm; A quality control module, used to perform anomaly detection and / or data repair on the depth feature data using a preset quality control algorithm to obtain quality controlled data; the quality control algorithm includes an anomaly detection algorithm and / or a data repair algorithm; An analysis and processing module, used to perform parallel processing and analysis on the quality-controlled data using a preset analysis and processing algorithm to obtain rail transit data analysis results; the analysis and processing algorithm includes a combination of one or more of a dynamic load balancing algorithm, a task similarity analysis algorithm, and an elastic scaling algorithm; Among them, the adaptive feature engineering algorithm, the quality control algorithm and the analysis and processing algorithm are dynamically optimized by means of deep transfer learning.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the program, the rail transit data analysis method based on adaptive feature engineering as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the rail transit data analysis method based on adaptive feature engineering as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the rail transit data analysis method based on adaptive feature engineering as described in any one of claims 1 to 6 is implemented.

Images