Citizen card consumption data analysis system and method based on cloud platform

Abstract

The invention relates to the technical field of citizen card data analysis, and particularly discloses a citizen card consumption data analysis system and method based on a cloud platform, and the method comprises the steps: carrying out the statistics of the morning peak passenger flow based on the subway arrival card swiping data of a subway station; the method comprises the following steps: performing data arrangement and periodic decomposition on historical data of early peak passenger flow volume of a subway station by using a data processing and prediction technology based on artificial intelligence, and then performing time sequence feature extraction on time subsequences of the early peak passenger flow volume; therefore, the early peak passenger flow volume is intelligently predicted according to the time sequence significant aggregation representation between the early peak passenger flow volume time sequence characteristics, and then whether a passenger flow volume abnormity alarm prompt is generated or not is judged. Therefore, significant features in different periods can be effectively decomposed by using rich historical data, and complex periodic and nonlinear changes are captured, so that the accuracy and stability of a prediction model are improved, and powerful support is provided for intelligent traffic management of a city.

Description

Technical Field

[0001] The present application relates to the technical field of citizen card data analysis, and more specifically, to a citizen card consumption data analysis system and method based on a cloud platform. Background Art

[0002] With the acceleration of urbanization and the continuous growth of urban populations, public transportation systems face tremendous pressure and challenges. As an efficient and convenient mode of public transportation, subways play a vital role in urban commuting. During the morning rush hour, subway station passenger flow is high and fluctuates frequently, placing extremely high demands on operational management.

[0003] However, traditional data analysis often relies on simple statistical methods, such as calculating averages, maximums, and minimums, failing to deeply explore the complex patterns and regularities hidden within the data. For data such as morning peak passenger flow, which exhibits distinct periodicity and dynamic variations, effective periodic decomposition and time series feature extraction are difficult to perform, making it difficult to accurately capture the underlying mechanisms of passenger flow over time. Furthermore, because traditional methods fail to fully utilize the rich information in historical data and fail to comprehensively consider the impact of multiple factors (such as weather and holidays) on morning peak passenger flow, their predictions are less accurate. This not only impacts optimal resource allocation but can also lead to low operational efficiency and reduced service quality.

[0004] Therefore, a cloud-based platform-based citizen card consumption data analysis solution is desired. Summary of the Invention

[0005] This application provides a citizen card consumption data analysis system and method based on a cloud platform, which can use rich historical data to effectively decompose significant features in different periods, capture complex periodic and nonlinear changes, thereby improving the accuracy and stability of the prediction model, and providing strong support for the city's intelligent traffic management.

[0006] According to one aspect of the present application, a method for analyzing citizen card consumption data based on a cloud platform is provided, comprising: Obtain the subway card entry data for the first subway station during the morning rush hour; Uploading the subway card swiping data of the first subway station to the cloud platform; On the cloud platform, based on the subway card swiping data of the first subway station, counting the passenger flow of the first subway station during the morning rush hour; On the cloud platform, extracting historical data of morning peak passenger flow at the first subway station from a backend database; On the cloud platform, based on the historical data of the morning peak passenger flow of the first subway station, the morning peak passenger flow of the current day is predicted to obtain the predicted morning peak passenger flow of the first subway station, including: performing a passenger flow core feature aggregation analysis based on time series features on the historical data of the morning peak passenger flow of the first subway station to obtain the predicted morning peak passenger flow of the first subway station; On the cloud platform, based on a comparison between the morning peak passenger flow of the first subway station and the predicted morning peak passenger flow of the first subway station, it is determined whether to generate an abnormal passenger flow alarm prompt.

[0007] In the above-mentioned cloud platform-based citizen card consumption data analysis method, the historical data of the morning peak passenger flow of the first subway station is subjected to a passenger flow core feature aggregation analysis based on time series characteristics to obtain the predicted morning peak passenger flow of the first subway station, including: Performing data sorting and periodic sequence feature extraction on historical data of morning peak passenger flow at the first subway station to obtain multiple time subsequences of morning peak passenger flow; Performing passenger flow time series feature extraction on each of the multiple morning peak passenger flow time subsequences to obtain multiple morning peak passenger flow time series feature encoding vectors; Performing a passenger flow time series multi-period feature significant aggregation analysis on the multiple morning peak passenger flow time series feature coding vectors to obtain a morning peak passenger flow multi-period significant aggregation coding vector; Based on the multi-time series period significant aggregation coding vector of the morning peak passenger flow, the predicted morning peak passenger flow of the first subway station is obtained.

[0008] In the above-mentioned cloud-based citizen card consumption data analysis method, the historical data of the morning peak passenger flow of the first subway station is sorted and periodic sequence features are extracted to obtain multiple time subsequences of the morning peak passenger flow, including: The historical data of the morning peak passenger flow of the first subway station is sorted according to the time dimension to obtain a time series of the morning peak passenger flow; The time series of the morning peak passenger flow is processed using a singular spectrum decomposition algorithm to extract the periodic characteristics of the time series of the morning peak passenger flow to obtain the multiple time subsequences of the morning peak passenger flow.

[0009] In the above-mentioned cloud platform-based citizen card consumption data analysis method, passenger flow time series feature extraction is performed on each time subsequence of the multiple morning peak passenger flows to obtain multiple morning peak passenger flow time series feature coding vectors, including: using a Bi-GRU-based feature extractor to process each time subsequence of the multiple morning peak passenger flows to obtain the multiple morning peak passenger flows time series feature coding vectors.

[0010] In the above-mentioned cloud platform-based citizen card consumption data analysis method, a significant aggregation analysis of the multi-period characteristics of passenger flow time series is performed on the multiple morning peak passenger flow time series feature coding vectors to obtain a significant aggregation coding vector of the multi-period of morning peak passenger flow time series, including: Performing passenger flow information core coarse-grained aggregation on the multiple morning peak passenger flow time series feature coding vectors to obtain a morning peak passenger flow time series coarse-grained aggregation coding vector; Based on the morning peak passenger flow time series coarse-grained aggregated coding vector, gated explicit fine-grained compensation aggregation is performed on the multiple morning peak passenger flow time series feature coding vectors to obtain a passenger flow time series fine-grained compensation aggregated coding vector; The morning peak passenger flow time series coarse-grained aggregation coding vector and the passenger flow time series fine-grained compensation aggregation coding vector are interactively analyzed to obtain the morning peak passenger flow multi-time series period significant aggregation coding vector.

[0011] In the above-mentioned cloud platform-based citizen card consumption data analysis method, based on the morning peak passenger flow time series coarse-grained aggregated coding vector, the gated explicit fine-grained compensation aggregation is performed on the multiple morning peak passenger flow time series feature coding vectors to obtain the passenger flow time series fine-grained compensated aggregated coding vector, including: Calculating a core convergence compensation factor of each morning peak passenger flow time series feature coding vector in the multiple morning peak passenger flow time series feature coding vectors relative to the morning peak passenger flow time series coarse-grained convergence coding vector to obtain multiple passenger flow time series core convergence compensation factors; Performing gated explicit compensation on the plurality of passenger flow timing core convergence compensation factors to obtain a plurality of passenger flow timing core convergence compensation weight factors; The multiple passenger flow time series core aggregation compensation weight factors, the morning peak passenger flow time series coarse-grained aggregation coding vectors and the multiple morning peak passenger flow time series feature coding vectors are subjected to passenger flow time series fine-grained dynamic compensation aggregation to obtain the passenger flow time series fine-grained compensation aggregation coding vector.

[0012] In the above-mentioned cloud platform-based citizen card consumption data analysis method, the core convergence compensation factor of each morning peak passenger flow time series feature coding vector in the multiple morning peak passenger flow time series feature coding vectors relative to the morning peak passenger flow time series coarse-grained convergence coding vector is calculated to obtain multiple passenger flow time series core convergence compensation factors, including: Calculating a morning peak passenger flow time series difference coding vector between the morning peak passenger flow time series feature coding vector and the morning peak passenger flow time series coarse-grained aggregation coding vector; The morning peak passenger flow time series differential coding vector is multiplied by the passenger flow time series weight matrix, and then added to the passenger flow time series offset value position by position to obtain the morning peak passenger flow time series compensation vector; Multiplying the morning peak passenger flow time series compensation vector by the scoring weight vector to obtain a passenger flow time series kernel convergence compensation factor corresponding to the morning peak passenger flow time series feature coding vector; In response to the fact that the second norm of the enhanced morning peak passenger flow time series feature coding vector is less than the second norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector, a logarithmic function value obtained by adding a constant one to the ratio of the second norm of the enhanced morning peak passenger flow time series feature coding vector to the second norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector is taken as the passenger flow time series offset value; In response to the fact that the binary norm of the enhanced morning peak passenger flow time series feature coding vector is greater than or equal to the binary norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector, the ratio of the binary norm of the enhanced morning peak passenger flow time series feature coding vector and the binary norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector is used as the passenger flow time series bias value.

[0013] In the above-mentioned cloud platform-based citizen card consumption data analysis method, the predicted morning peak passenger flow of the first subway station is obtained based on the multi-time series period significant convergence coding vector of the morning peak passenger flow, including: inputting the multi-time series period significant convergence coding vector of the morning peak passenger flow into the morning peak passenger flow prediction module based on the RNN model to obtain the predicted morning peak passenger flow of the first subway station.

[0014] In the above-mentioned cloud platform-based citizen card consumption data analysis method, determining whether to generate an abnormal passenger flow alarm based on a comparison between the morning peak passenger flow at the first subway station and the predicted morning peak passenger flow at the first subway station includes: Calculating a ratio between the morning peak passenger flow of the first subway station and the predicted morning peak passenger flow of the first subway station as a passenger flow deviation rate; Based on the comparison between the passenger flow deviation rate and a preset threshold, it is determined whether to generate the passenger flow abnormality alarm prompt.

[0015] According to another aspect of the present application, a citizen card consumption data analysis system based on a cloud platform is also provided, comprising: The subway station card swiping data acquisition module is used to obtain the subway station card swiping data of the first subway station during the morning rush hour; A subway station card swiping data uploading module, used to upload the subway station card swiping data of the first subway station to the cloud platform; A morning peak passenger flow statistics module is used to count the morning peak passenger flow of the first subway station based on the subway card swiping data of the first subway station on the cloud platform; A passenger flow historical data extraction module extracts the historical data of the morning peak passenger flow of the first subway station from the backend database on the cloud platform; a morning peak passenger flow prediction module, configured to predict the morning peak passenger flow of the day on the cloud platform based on the historical data of the morning peak passenger flow of the first subway station to obtain the predicted morning peak passenger flow of the first subway station, wherein the morning peak passenger flow prediction module comprises: performing a passenger flow core feature aggregation analysis based on time series features on the historical data of the morning peak passenger flow of the first subway station to obtain the predicted morning peak passenger flow of the first subway station; The passenger flow abnormality alarm module is used to determine whether to generate a passenger flow abnormality alarm prompt on the cloud platform based on the comparison between the morning peak passenger flow of the first subway station and the predicted morning peak passenger flow of the first subway station.

[0016] The present application provides a citizen card consumption data analysis system and method based on a cloud platform, which obtains the subway card swiping data of the first subway station during the morning rush hour and uploads it to the cloud platform. In the cloud platform, first, based on the subway card swiping data of the subway station, the morning rush hour passenger flow is counted and the historical data of the morning rush hour passenger flow is extracted. Then, the data processing and prediction technology based on artificial intelligence is used to sort and periodically decompose the historical data of the morning rush hour passenger flow of the subway station. Then, the time series feature extraction is performed on the time subseries of each morning rush hour passenger flow, so as to intelligently predict the morning rush hour passenger flow based on the time series significant aggregation representation between the time series features of each morning rush hour passenger flow, and judge whether to generate a passenger flow abnormality alarm prompt based on the comparison between the ratio between the predicted value and the actual value and the preset threshold. Compared with the traditional method, the present application can effectively decompose the significant features in different periods by using rich historical data, capture complex periodic and nonlinear changes, thereby improving the accuracy and stability of the prediction model and providing strong support for the intelligent traffic management of the city. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings of the embodiments of the present application. Obviously, the drawings described below only relate to some embodiments of the present application, and are not intended to limit the present application.

[0018] Figure 1 This is a schematic flow chart of the citizen card consumption data analysis method based on the cloud platform in an embodiment of the present application.

[0019] Figure 2 This is a schematic flowchart of step S5 in the citizen card consumption data analysis method based on the cloud platform in an embodiment of the present application.

[0020] Figure 3 This is a schematic flowchart of step S51 in the citizen card consumption data analysis method based on the cloud platform in an embodiment of the present application.

[0021] Figure 4 This is a schematic flowchart of step S53 in the citizen card consumption data analysis method based on the cloud platform in an embodiment of the present application.

[0022] Figure 5 This is a schematic flowchart of step S532 in the citizen card consumption data analysis method based on the cloud platform in an embodiment of the present application.

[0023] Figure 6 This is a schematic flowchart of step S6 in the citizen card consumption data analysis method based on the cloud platform in an embodiment of the present application.

[0024] Figure 7 This is a schematic block diagram of a citizen card consumption data analysis system based on a cloud platform according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts also fall within the scope of protection of this application.

[0026] Based on the problems in the above background, this application proposes a citizen card consumption data analysis method based on a cloud platform, which obtains the subway card swiping data of the first subway station during the morning rush hour and uploads it to the cloud platform. In the cloud platform, first, based on the subway card swiping data of the subway station, the morning rush hour passenger flow is counted, and the historical data of the morning rush hour passenger flow is extracted. Then, the data processing and prediction technology based on artificial intelligence is used to sort and periodically decompose the historical data of the morning rush hour passenger flow of the subway station. Then, the time series features of each morning rush hour passenger flow are extracted, so as to intelligently predict the morning rush hour passenger flow based on the time series significant aggregation representation between the time series features of each morning rush hour passenger flow, and judge whether to generate a passenger flow abnormality alarm based on the comparison between the ratio between the predicted value and the actual value and the preset threshold. Compared with the traditional method, this application can effectively decompose the significant features in different periods by using rich historical data, capture complex periodic and nonlinear changes, thereby improving the accuracy and stability of the prediction model, and providing strong support for the intelligent traffic management of the city.

[0027] Figure 1 This is a schematic flow chart of a method for analyzing citizen card consumption data based on a cloud platform according to an embodiment of the present application. Figure 1 As shown, the citizen card consumption data analysis method based on the cloud platform includes: S1, obtaining the subway card swiping data of the first subway station during the morning rush hour; S2, uploading the subway card swiping data of the first subway station to the cloud platform; S3, on the cloud platform, based on the subway card swiping data of the first subway station, counting the morning rush hour passenger flow of the first subway station; S4, on the cloud platform, extracting the historical data of the morning rush hour passenger flow of the first subway station from the background database; S5, on the cloud platform, predicting the morning rush hour passenger flow of the day based on the historical data of the morning rush hour passenger flow of the first subway station to obtain the predicted morning rush hour passenger flow of the first subway station; S6, on the cloud platform, determining whether to generate a passenger flow abnormality alarm prompt based on the comparison between the morning rush hour passenger flow of the first subway station and the predicted morning rush hour passenger flow of the first subway station.

[0028] For example, in step S1, subway card entry data for the first subway station during the morning rush hour is obtained. It should be understood that the morning rush hour is typically one of the busiest times for urban traffic. For the subway system, passenger flow during this period is large and fluctuates frequently. Therefore, accurately understanding passenger entry and exit patterns during this period is crucial for optimizing operational management and improving service quality. Specifically, this data can help identify peak passenger flow, assess congestion levels, identify potential safety hazards, and provide a basis for formulating emergency measures.

[0029] In a specific example of this application, the citizen card, as an important payment tool in the public transportation system, records the travel information of passengers when they enter and exit the subway station. Every time a passenger uses a citizen card to pass through the subway gate, the system automatically records the time, place, card number and other information of the card swipe. These data are stored in the background database of the subway system, forming a huge card swipe record library. In order to focus on data collection during the morning rush hour, a fixed time window can be set, such as between 7 and 9 in the morning on weekdays. All card swipe events that occur during this time period will be specially marked and extracted.

[0030] For example, in step S2, the subway card entry data for the first subway station is uploaded to the cloud platform. It should be understood that using a cloud platform for processing offers multiple advantages, the first of which is its powerful computing power and flexible scalability. The cloud platform can easily process massive data sets and apply complex algorithmic models to mine deep insights from the data without worrying about local hardware resource limitations. Furthermore, the cloud platform provides a high degree of security and reliability, ensuring the security of data during transmission and storage. It also supports real-time data analysis, enabling rapid response to changes and providing timely prediction results and anomaly alerts, which is crucial for dynamically adjusting subway operation strategies.

[0031] In a specific example of the present application, uploading the subway card entry data of the first subway station to the cloud platform includes: first, it is necessary to establish a stable and reliable data transmission channel, which can usually be achieved through an API interface. Specifically, the background database of the subway system will regularly generate data files containing card swiping records during the morning rush hour, and these files will then be formatted into a standard format suitable for transmission, such as JSON or CSV. Next, scripts are written in a programming language (such as Python) or existing ETL tools are used to automatically perform data extraction, conversion, and loading operations according to predetermined time intervals or trigger conditions. In this process, data encryption technology is used to protect data privacy and security to ensure that no leakage occurs during transmission. Once the data arrives at the cloud platform, it will be stored in a specially designed database so that subsequent data processing and analysis tasks can proceed smoothly. In this way, not only is the efficient uploading of data achieved, but it also lays a solid foundation for subsequent in-depth analysis based on the powerful computing power of the cloud platform.

[0032] For example, in step S3, the cloud platform calculates the morning peak passenger flow at the first subway station based on the subway card entry data. It should be understood that morning peak passenger flow refers to the number of passengers passing through a specific location (such as a subway station or bus stop) during the morning rush hour, typically between 7:00 and 9:00 a.m. on weekdays. Subway card entry data is a true record of passengers' actual station entry behavior and is the most direct way to obtain morning peak passenger flow. By understanding morning peak passenger flow, the congestion level of platforms and trains can be assessed, allowing appropriate safety measures to be implemented in advance. For example, when passenger flow exceeds a certain threshold, flow control measures can be implemented to control the number of people entering the platform, prevent safety accidents caused by overcrowding, and ensure the personal safety of passengers.

[0033] For example, in step S4, the cloud platform extracts historical data on the morning peak passenger flow of the first subway station from the backend database. It should be understood that, considering that the historical data on the morning peak passenger flow contains detailed passenger flow information of the station during the past morning peak hours, these data accumulate over time and are rich in details and continuity. It records the passenger flow conditions on different dates, seasons, and weather conditions, providing a sufficient data basis for subsequent analysis. By studying these data, we can fully understand the characteristics and changes of the morning peak passenger flow of the station. The current morning peak passenger flow situation is closely related to the historical data. Historical data can provide a reference and comparison for the current passenger flow analysis, helping analysts to better understand the current passenger flow status. For example, by comparing historical data for the same period, we can determine whether the current passenger flow is normal and whether there are abnormal fluctuations, so as to more accurately predict the passenger flow of the station during the future morning peak hours.

[0034] Exemplarily, in step S5, on the cloud platform, the morning peak passenger flow of the day is predicted based on the historical data of the morning peak passenger flow of the first subway station to obtain the predicted morning peak passenger flow of the first subway station, including: performing a passenger flow core feature aggregation analysis based on time series features on the historical data of the morning peak passenger flow of the first subway station to obtain the predicted morning peak passenger flow of the first subway station.

[0035] In one embodiment, Figure 2As shown, the historical data of the morning peak passenger flow of the first subway station is subjected to a passenger flow core feature aggregation analysis based on time series features to obtain the predicted morning peak passenger flow of the first subway station, including: S51, data sorting and periodic sequence feature extraction of the historical data of the morning peak passenger flow of the first subway station to obtain multiple time subsequences of morning peak passenger flow; S52, passenger flow time series feature extraction of each time subsequence of the multiple morning peak passenger flow time subsequences to obtain multiple morning peak passenger flow time series feature coding vectors; S53, passenger flow time series multi-period feature significant aggregation analysis of the multiple morning peak passenger flow time series feature coding vectors to obtain a morning peak passenger flow multi-time series period significant aggregation coding vector; S54, based on the morning peak passenger flow multi-time series period significant aggregation coding vector, the predicted morning peak passenger flow of the first subway station is obtained.

[0036] For example, Figure 3 As shown, in step S51, the historical data of the morning peak passenger flow of the first subway station is sorted and periodic sequence features are extracted to obtain multiple time subsequences of the morning peak passenger flow, including: S511, the historical data of the morning peak passenger flow of the first subway station is sorted according to the time dimension to obtain the time series of the morning peak passenger flow; S512, the time series of the morning peak passenger flow is processed using a singular spectrum decomposition algorithm to extract the periodic features of the time series of the morning peak passenger flow to obtain the multiple time subsequences of the morning peak passenger flow.

[0037] Specifically, in step S511, considering that passenger flow data inherently has distinct temporal attributes, changes in morning peak passenger flow at different time points reflect its inherent patterns and trends, such as periodic changes (daily, hourly, etc.), trend changes (increasing or decreasing trends), etc. Therefore, in the technical solution of this application, the historical data of morning peak passenger flow at the first subway station is sorted according to the time dimension to obtain a time series of morning peak passenger flow. This can better demonstrate the inherent logic and change patterns of the data, thereby more accurately understanding the dynamic changes in morning peak passenger flow.

[0038] Specifically, in step S512, considering that the time series of morning peak passenger flow often contains multiple complex components, such as long-term trends, seasonal cycles, random fluctuations, and possible outliers, different periodic features have different temporal characteristics. Therefore, in order to decompose the time series at different time scales and extract short-term, medium-term, and long-term trends and periodic components, the present application processes the time series of morning peak passenger flow using a singular spectrum decomposition algorithm to extract the periodic characteristics of the time series of morning peak passenger flow, thereby obtaining multiple time subseries of morning peak passenger flow. It should be understood that the singular spectrum decomposition algorithm is a non-parametric time series analysis method based on matrix decomposition. It converts the time series into a trajectory matrix and then performs singular value decomposition (SVD) on the matrix to decompose the time series into different components. These components can represent characteristics such as trend, cycle, seasonality, and noise in the time series. The singular spectrum decomposition algorithm can effectively decompose these components of different properties, thereby more clearly understanding the structure of the time series. This allows for accurate identification of cyclical patterns in morning peak passenger flow, such as similar daily variations in morning peak passenger flow or weekly differences in morning peak passenger flow between weekdays and weekends. This helps subway operations departments predict overall future passenger flow trends, enabling them to develop more targeted operational plans and rationally arrange train schedules and departure intervals.

[0039] In a specific example of this application, suppose there is a subway station A with entry card swipe data recorded during the morning rush hour (e.g., 7:00 AM to 9:00 AM) for 30 consecutive weekdays. This data forms a time series T with 30 data points. First, a trajectory matrix X is constructed. Assuming a window length L of 10 days, this means that each 10 consecutive days of data is considered a subsequence and a column in the trajectory matrix. Therefore, for 30 days of data, a 10x21 matrix X is obtained (because days 1 to 10 form the first column, days 2 to 11 form the second column, and so on, up to the 21st column). Each row of this matrix represents the value at a specific moment in the time series, while each column represents the data within a sliding window. Next, a singular value decomposition (SVD) is performed on the trajectory matrix X: X = UΣV^T, where U and V are orthogonal matrices and Σ is a diagonal matrix whose diagonal elements are the singular values ​​of X. Singular values ​​reflect the importance of different components. Components corresponding to larger singular values ​​usually contain the main trend or periodic information, while smaller singular values ​​may correspond to noise. Then, based on the singular values ​​and their corresponding left and right singular vectors, different time series components can be reconstructed. Specifically, the first few largest singular values ​​(for example, the first three) are selected, and new matrices are reconstructed based on the left and right singular vectors corresponding to these singular values. These matrices are then converted back into time series form. In this way, several time subsequences can be obtained, each of which represents a different characteristic of the original time series, such as long-term trends, seasonal fluctuations, or random noise.

[0040] Exemplarily, in step S52, passenger flow time series feature extraction is performed on each of the multiple morning peak passenger flow time subsequences to obtain multiple morning peak passenger flow time series feature encoding vectors. It should be understood that the multiple morning peak passenger flow time subsequences obtained through singular spectrum decomposition each represent different components, such as trend components, periodic components, and noise components. Each subsequence has unique characteristics and variation patterns. To more comprehensively and meticulously capture the various information of the morning peak passenger flow time series, the technical solution of the present application performs passenger flow time series feature extraction on each of the multiple morning peak passenger flow time subsequences to fully exploit the information contained in these different subsequences, thereby obtaining multiple morning peak passenger flow time series feature encoding vectors. This allows for a deeper understanding of the inherent characteristics of morning peak passenger flow. For example, the feature encoding vectors of the trend subsequences can reveal the specific characteristics of the long-term growth or decline trend of passenger flow; the feature encoding vectors of the periodic subsequences can reveal information such as the amplitude and frequency of periodic changes in passenger flow.

[0041] In particular, in a specific example of the present application, passenger flow time series feature extraction is performed on each time subsequence of the multiple time subsequences of morning peak passenger flow to obtain multiple morning peak passenger flow time series feature encoding vectors, including: using a Bi-GRU-based feature extractor to process each time subsequence of the multiple time subsequences of morning peak passenger flow to obtain the multiple morning peak passenger flow time series feature encoding vectors. That is to say, in the time subsequence of morning peak passenger flow, the passenger flow at each moment is not only related to the past state, but may also be affected by the future state. Bi-GRU consists of two GRUs in opposite directions, one processing data from the start end to the end end of the sequence, and the other processing data from the end end to the start end. This can capture both past and future information in the time series at the same time, and can more comprehensively understand the contextual information of the time series compared to unidirectional recurrent neural networks (such as ordinary GRU or LSTM). For example, when analyzing passenger flow during the morning rush hour, the passenger flow at a certain moment may be affected by the travel habits of passengers in the previous period and subsequent upcoming events (such as the start time of a large-scale event). Bi-GRU can better capture these complex dependencies to better learn the changing patterns of passenger flow and make more accurate predictions.

[0042] Exemplarily, in step S53, the multiple morning peak passenger flow time series feature coding vectors are subjected to a significant aggregation analysis of passenger flow time series multi-period features to obtain a morning peak passenger flow multi-period significant aggregation coding vector. It should be understood that, given that each morning peak passenger flow time series feature coding vector extracts feature information from different angles or subsequences, it each contains some time series features related to passenger flow. However, these features may be relatively dispersed, making it difficult to form a more comprehensive and representative representation. Furthermore, morning peak passenger flow typically exhibits characteristics across multiple time periods, such as daily periodicity, weekly periodicity, and possible seasonal periodicity. A single time series feature coding vector may only reflect the periodic characteristics of one or a few aspects. Therefore, in order to comprehensively consider information from multiple periods, more completely capture the multi-period variation patterns of morning peak passenger flow, and provide a more comprehensive and comprehensive overall variation trend of passenger flow, the present application performs a significant aggregation analysis of passenger flow time series multi-period features on the multiple morning peak passenger flow time series feature coding vectors to obtain a morning peak passenger flow multi-period significant aggregation coding vector. In particular, this method grasps the overall passenger flow change trend through coarse-grained aggregation, combines the kernel aggregation compensation factor with the fine-grained dynamic compensation mechanism, and models the local details of each site node with high fidelity. Finally, it integrates complex features to generate a significant convergence representation of multiple time series periods of morning peak passenger flow with both global and local characteristics, providing a better basis for passenger flow prediction.

[0043] In one embodiment, Figure 4 As shown, the multiple morning peak passenger flow time series feature coding vectors are subjected to passenger flow time series multi-period feature significant aggregation analysis to obtain the morning peak passenger flow multi-period significant aggregation coding vector, including: S531, passenger flow information core coarse-grained aggregation is performed on the multiple morning peak passenger flow time series feature coding vectors to obtain the morning peak passenger flow time series coarse-grained aggregation coding vector; S532, based on the morning peak passenger flow time series coarse-grained aggregation coding vector, the multiple morning peak passenger flow time series feature coding vectors are subjected to gated explicit fine-grained compensation aggregation to obtain the passenger flow time series fine-grained compensation aggregation coding vector; S533, the morning peak passenger flow time series coarse-grained aggregation coding vector and the passenger flow time series fine-grained compensation aggregation coding vector are interactively analyzed to obtain the morning peak passenger flow multi-period significant aggregation coding vector.

[0044] Specifically, in step S531, the passenger flow information kernel coarse-grained aggregation is performed on the multiple morning peak passenger flow time series feature coding vectors to obtain a morning peak passenger flow time series coarse-grained aggregation coding vector, which is expressed as follows: in, are the multiple morning peak passenger flow time series feature coding vectors, , , and are the first, second, and third of the multiple morning peak passenger flow time series feature coding vectors. and A time series feature coding vector of passenger flow during the morning peak period, and Take The maximum and minimum values ​​of It is the first value among the multiple morning peak passenger flow time series characteristics convergence values. The convergence value of the time series characteristics of the morning peak passenger flow, is the normalization function, It is the first value among multiple normalized multiple morning peak passenger flow time series feature aggregation values. Normalized multiple morning peak passenger flow time series feature aggregation values, is the The number of vectors in It is the coarse-grained aggregation coding vector of the morning peak passenger flow time series.

[0045] Specifically, the process of performing coarse-grained aggregation of multiple morning peak passenger flow time series feature encoding vectors using passenger flow information kernels to obtain a coarse-grained aggregated encoding vector for the morning peak passenger flow time series is essentially an attempt to extract a global, generalized representation from a large number of local details. This process is crucial for understanding and predicting morning peak passenger flow at subway stations. Specifically, multiple morning peak passenger flow time series feature encoding vectors may contain complex patterns and trends, such as cyclical fluctuations, seasonal variations, and the impact of unexpected events. However, directly analyzing these morning peak passenger flow time series feature encoding vectors can be difficult due to data complexity and noise. Therefore, a method is needed to simplify this data and extract the most important information. In this step, these feature encoding vectors are compressed into a more concise representation—the coarse-grained aggregated encoding vector for the morning peak passenger flow time series—using a coarse-grained aggregation network. This aggregated encoding vector can be considered a global description of the entire time series collection, capturing the main trends and patterns across all time series.

[0046] Specifically, if Figure 5 As shown, in step S532, based on the morning peak passenger flow time series coarse-grained aggregation coding vector, the multiple morning peak passenger flow time series feature coding vectors are gated explicitly fine-grained compensation aggregation to obtain a passenger flow time series fine-grained compensation aggregation coding vector, including: S5321, calculating the core convergence compensation factor of each morning peak passenger flow time series feature coding vector in the multiple morning peak passenger flow time series feature coding vectors relative to the morning peak passenger flow time series coarse-grained aggregation coding vector to obtain multiple passenger flow time series core convergence compensation factors; S5322, performing gated explicit compensation on the multiple passenger flow time series core convergence compensation factors to obtain multiple passenger flow time series core convergence compensation weight factors; S5323, performing passenger flow time series fine-grained dynamic compensation aggregation on the multiple passenger flow time series core convergence compensation weight factors, the morning peak passenger flow time series coarse-grained aggregation coding vector and the multiple morning peak passenger flow time series feature coding vectors to obtain the passenger flow time series fine-grained compensation aggregation coding vector.

[0047] In one embodiment, in step S5321, the core convergence compensation factor of each morning peak passenger flow time series feature coding vector in the multiple morning peak passenger flow time series feature coding vectors relative to the morning peak passenger flow time series coarse-grained convergence coding vector is calculated to obtain multiple passenger flow time series core convergence compensation factors, including: calculating the morning peak passenger flow time series difference coding vector between the morning peak passenger flow time series feature coding vector and the morning peak passenger flow time series coarse-grained convergence coding vector, which is expressed by the formula: in, is point convolutional coding, is the activation function, and They are and The corresponding weight matrix, yes The corresponding enhanced morning peak passenger flow time series feature coding vector, is the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector, It is subtracted by position point, To take the absolute value operation, for and The time series difference encoding vector of the morning peak passenger flow.

[0048] The morning peak passenger flow time series differential coding vector is multiplied by the passenger flow time series weight matrix and then added to the passenger flow time series offset value position by position to obtain the morning peak passenger flow time series compensation vector; the morning peak passenger flow time series compensation vector is multiplied by the scoring weight vector to obtain the passenger flow time series core convergence compensation factor corresponding to the morning peak passenger flow time series feature coding vector, which is expressed as follows: in, yes The corresponding passenger flow time series weight matrix, is matrix multiplication, for The corresponding passenger flow time series offset value, for The corresponding scoring weight vector, It is the compensation factor of multiple passenger flow time series core convergence The passenger flow time series kernel convergence compensation factor.

[0049] Specifically, the differential encoding vector is first calculated between each morning peak passenger flow time series feature encoding vector and the coarse-grained aggregate encoding vector. This differential encoding vector effectively reflects the degree of deviation between each day's passenger flow pattern and the overall trend. Specifically, if a day's passenger flow is significantly above average, the differential encoding vector for that day will display a large positive value; conversely, if it is below average, it will display a negative value. This method accurately quantifies the uniqueness of each day relative to the overall trend. Next, these morning peak passenger flow time series differential encoding vectors are multiplied by a predefined passenger flow time series weight matrix. This weight matrix weights the differential encoding vectors based on the importance of different dimensions. For example, some dimensions (such as holiday impact) may have a greater impact on passenger flow than others (such as weather conditions) and therefore require higher weights. The adjusted differential encoding vectors are then added position-by-position to a passenger flow time series offset value to generate a morning peak passenger flow time series compensation vector. This compensation vector not only takes into account the importance of each dimension but also incorporates fine-tuning using the offset value to ensure more accurate compensation. Finally, the morning peak passenger flow time series compensation vector is multiplied by the scoring weight vector to generate the passenger flow time series kernel convergence compensation factor corresponding to each morning peak passenger flow time series feature encoding vector. This compensation factor essentially describes the correction rule for each time series relative to the overall trend. It helps to highlight time series with significant local features while preserving the global trend. In particular, those skilled in the art should be aware that the weight matrix and scoring weight vector in the model are automatically learned through training. For example, the backpropagation algorithm is used to optimize the weight matrix and scoring weight vector to minimize the prediction error.

[0050] In which, in response to the fact that the second norm of the enhanced morning peak passenger flow time series feature coding vector is less than the second norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector, the ratio of the second norm of the enhanced morning peak passenger flow time series feature coding vector to the second norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector is added with a constant one, and the logarithm obtained by taking the logarithm with base 2 is used as the passenger flow time series offset value; in response to the fact that the second norm of the enhanced morning peak passenger flow time series feature coding vector is greater than or equal to the second norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector, the ratio of the second norm of the enhanced morning peak passenger flow time series feature coding vector to the second norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector is used as the passenger flow time series offset value, which is expressed by the formula: in, To calculate the two norm of a vector, is the base 2 logarithmic function value.

[0051] Here, for the compensation of the deviation between the morning peak passenger flow time series feature coding vector and the morning peak passenger flow time series coarse-grained aggregation coding vector, the performance deviation of the kernel convergence strategy as a scenario strategy can be measured by quantifying the regret metric based on the information kernel compression assumption in the kernel convergence decision-making process, that is, the game counterfactual regret value. Specifically, the norm of the vector is used to provide a normalized decision point loss description based on the policy action, that is, the vector norm representation of the morning peak passenger flow time series feature coding vector and the morning peak passenger flow time series coarse-grained aggregation coding vector. Then, for the possible differences in the vector distribution action game scenario, the compensation rule of the node personalized information is modified by the degree of information distribution of the regret value and the relative distribution amplitude of the regret value, so that the node personalized information is considered as the action not taken in the decision, and the biased compensation is performed based on the potential benefit of the information kernel convergence assumption.

[0052] Specifically, in step S5322, gated explicit compensation is performed on the multiple passenger flow timing core convergence compensation factors to obtain multiple passenger flow timing core convergence compensation weight factors, which are expressed as follows: in, For Perform gate compensation, is the preset compensation threshold, It is the compensation weight factor of multiple passenger flow time series core convergence The passenger flow time series kernel aggregation compensation weight factor.

[0053] It should be understood that the process of performing gated explicit compensation on multiple passenger flow time series kernel convergence compensation factors to obtain multiple passenger flow time series kernel convergence compensation weight factors actually involves a dynamic adjustment mechanism to refine the local feature correction of each time series. This process is crucial for improving the accuracy and robustness of the morning peak passenger flow prediction model. In this step, gated explicit compensation is performed on multiple passenger flow time series kernel convergence compensation factors to generate multiple passenger flow time series kernel convergence compensation weight factors. Specifically, gating functions are widely used in deep neural networks. Their core purpose is to dynamically select information based on nonlinear constraints. Here, the gated explicit compensation mechanism uses a learnable parameterized gating function to control the effect of the compensation factors. The effectiveness of this gated explicit compensation is that it ensures that only truly important local features are highlighted, while irrelevant or redundant information is appropriately suppressed.

[0054] Specifically, in step S5323, the multiple passenger flow time series core aggregation compensation weight factors, the morning peak passenger flow time series coarse-grained aggregation coding vector, and the multiple morning peak passenger flow time series feature coding vectors are subjected to passenger flow time series fine-grained dynamic compensation aggregation to obtain the passenger flow time series fine-grained compensation aggregation coding vector, which is expressed as follows: in, It is the passenger flow time series fine-grained compensation aggregation coding vector.

[0055] It should be understood that the process of performing fine-grained dynamic compensation aggregation of passenger flow time series by combining multiple passenger flow time series kernel aggregation compensation weight factors, coarse-grained aggregation encoding vectors of morning peak passenger flow time series, and multiple morning peak passenger flow time series feature encoding vectors actually uses a fine-tuning mechanism to integrate global trends and local details to generate a more accurate and comprehensive representation—the fine-grained compensation aggregation encoding vector of the passenger flow time series. This process is crucial for improving the accuracy and robustness of the morning peak passenger flow prediction model. In this step, these compensation weight factors are combined with the coarse-grained aggregation encoding vector of the morning peak passenger flow time series and multiple morning peak passenger flow time series feature encoding vectors to perform fine-grained dynamic compensation aggregation. Specifically, this process is similar to the attention mechanism, which allows the model to achieve coordinated feature representation at the global and local levels.

[0056] Specifically, in step S533, the morning peak passenger flow time series coarse-grained aggregation coding vector and the passenger flow time series fine-grained compensation aggregation coding vector are interactively analyzed to obtain the morning peak passenger flow multi-time series period significant aggregation coding vector, which is expressed as follows: in, and is the weighted hyperparameter, It is the encoding vector of the significant aggregation of multiple time series periods of the morning peak passenger flow. Here, it should be understood that the weighted hyperparameters are initialized based on historical experiments or domain knowledge, and then adjusted on the validation set through grid search (GridSearch) or Bayesian optimization (Bayesian Optimization) to balance the importance of coarse-grained and fine-grained features. It should be understood that the interactive analysis of the coarse-grained aggregated encoding vector of the morning peak passenger flow time series and the fine-grained compensated aggregated encoding vector of the passenger flow time series to obtain the significant aggregated encoding vector for multiple time series periods of the morning peak passenger flow actually integrates global trends and local details to produce a more comprehensive and accurate representation. This process is crucial for improving the accuracy and robustness of the morning peak passenger flow prediction model. The coarse-grained aggregated encoding vector of the morning peak passenger flow time series provides the overall trend over the entire time period. For example, in a month's worth of data, it can capture the difference between weekdays and weekends, or long-term growth trends. This global perspective helps understand overall patterns of change and identify cyclical fluctuations and other macro trends. The fine-grained compensated aggregated encoding vector of the passenger flow time series focuses on capturing local details and anomalies within each time series. For example, in the above scenario, some days see significant increases in passenger flow due to special events (such as large concerts or sporting events), while other days see decreases due to weather conditions (such as heavy rain or snow). These local fluctuations may be masked by the global trend, but they are crucial for accurate forecasting and decision-making. During the interaction analysis, these two vectors are combined using a weighting hyperparameter to balance the importance of global and local information.

[0057] Exemplarily, in step S54, based on the morning peak passenger flow multi-period significantly converged encoding vector, the predicted morning peak passenger flow for the first subway station is obtained, including: inputting the morning peak passenger flow multi-period significantly converged encoding vector into a morning peak passenger flow prediction module based on an RNN model to obtain the predicted morning peak passenger flow for the first subway station. In other words, the morning peak passenger flow multi-period significantly converged encoding vector, obtained by periodically significantly convergently performing prediction processing, is used to intelligently obtain the predicted morning peak passenger flow for the first subway station. It should be understood that the morning peak passenger flow multi-period significantly converged encoding vector contains rich time series feature information, encompassing the changing patterns of multiple time periods as well as global and local features. Due to its unique recurrent structure, the RNN model is capable of processing data with time series properties. Through memory units, it can capture long-term dependencies in the sequence and automatically learn the patterns and regularities hidden in the morning peak passenger flow multi-period significantly converged encoding vector. Through training on historical data, the RNN model can establish a mapping relationship between the input encoding vector and actual morning peak passenger flow, thereby predicting future passenger flow.

[0058] For example, Figure 6As shown, in step S6, on the cloud platform, based on the comparison between the morning peak passenger flow of the first subway station and the predicted morning peak passenger flow of the first subway station, it is determined whether to generate an abnormal passenger flow alarm prompt, including: S61, calculating the ratio between the morning peak passenger flow of the first subway station and the predicted morning peak passenger flow of the first subway station as the passenger flow deviation rate; S62, based on the comparison between the passenger flow deviation rate and a preset threshold, determining whether to generate the abnormal passenger flow alarm prompt.

[0059] Specifically, in step S61, the ratio between the morning peak passenger flow at the first subway station and the predicted morning peak passenger flow at the first subway station is calculated as the passenger flow deviation rate. It should be understood that there is often a discrepancy between the actual and predicted morning peak passenger flow. By calculating the ratio (passenger flow deviation rate), this discrepancy can be quantified using a specific numerical value. Compared to simply comparing two numerical values, the ratio can more intuitively reflect the degree of deviation between the predicted and actual values. The passenger flow deviation rate can help subway operators promptly detect abnormalities in morning peak passenger flow.

[0060] Specifically, in step S62, a determination is made as to whether to generate an abnormal passenger flow alert based on a comparison between the passenger flow deviation rate and a preset threshold. Specifically, the preset threshold here is 20%. Accordingly, in subway operations, stable passenger flow is crucial to ensuring operational efficiency and passenger experience. The passenger flow deviation rate intuitively reflects the degree to which actual morning peak passenger flow deviates from the predicted value. The preset threshold is determined based on a combination of factors, including historical operational data, passenger flow patterns, and operational experience, and serves as a quantitative definition of the range of passenger flow fluctuations under normal operating conditions. By comparing this threshold with the preset threshold, it is possible to quickly determine whether the current passenger flow situation deviates from the normal range. For example, if the deviation rate suddenly rises significantly, far exceeding the preset threshold, it is highly likely to indicate an abnormal situation, such as a large-scale event causing a large number of passengers to travel together, or factors such as line failures and inclement weather causing some passengers to change their travel methods and flock to a certain station. Upon receiving an abnormality alert, the operations department can promptly adjust its operational strategy based on the actual situation, such as temporarily deploying additional shuttle buses or extending operating hours, to alleviate congestion and ensure the normal operation of the subway.

[0061] In summary, according to the embodiment of the present application, the cloud platform-based citizen card consumption data analysis method is explained, which can use rich historical data to effectively decompose the significant features in different periods, capture complex periodic and nonlinear changes, thereby improving the accuracy and stability of the prediction model, and providing strong support for the city's intelligent traffic management.

[0062] Figure 7This is a schematic block diagram of a citizen card consumption data analysis system based on a cloud platform according to an embodiment of the present application. Figure 7 As shown, the citizen card consumption data analysis system 700 based on the cloud platform includes: a subway station card swiping data acquisition module 710, which is used to obtain the subway station card swiping data of the first subway station during the morning rush hour; a subway station card swiping data upload module 720, which is used to upload the subway station card swiping data of the first subway station to the cloud platform; a morning rush hour passenger flow statistics module 730, which is used to count the morning rush hour passenger flow of the first subway station based on the subway station card swiping data of the first subway station on the cloud platform; a passenger flow history data extraction module 740, which extracts the morning rush hour passenger flow history data of the first subway station from the background database on the cloud platform; the morning rush hour passenger flow The prediction module 750 is used to predict the morning peak passenger flow of the day on the cloud platform based on the historical data of the morning peak passenger flow of the first subway station to obtain the predicted morning peak passenger flow of the first subway station, wherein the morning peak passenger flow prediction module includes: performing a passenger flow core feature aggregation analysis based on time series features on the historical data of the morning peak passenger flow of the first subway station to obtain the predicted morning peak passenger flow of the first subway station; the passenger flow abnormality alarm module 760 is used to determine whether to generate a passenger flow abnormality alarm prompt on the cloud platform based on the comparison between the morning peak passenger flow of the first subway station and the predicted morning peak passenger flow of the first subway station.

[0063] Here, those skilled in the art will understand that the specific operations of each module in the above-mentioned citizen card consumption data analysis system based on the cloud platform have been referred to above. Figures 1 to 5 The description of the citizen card consumption data analysis method based on the cloud platform has been introduced in detail, and therefore, its repeated description will be omitted.

[0064] Finally, it should be noted that the embodiments described above are only some of the embodiments of the present invention, not all of them. The detailed description of the embodiments of the present invention is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are also within the scope of protection of the present invention.

Claims

1. A citizen card consumption data analysis method based on a cloud platform, characterized in that: include: Obtain the subway card entry data for the first subway station during the morning rush hour; Uploading the subway card swiping data of the first subway station to the cloud platform; On the cloud platform, based on the subway card swiping data of the first subway station, counting the passenger flow of the first subway station during the morning rush hour; On the cloud platform, extracting historical data of morning peak passenger flow at the first subway station from a backend database; On the cloud platform, based on the historical data of the morning peak passenger flow of the first subway station, the morning peak passenger flow of the current day is predicted to obtain the predicted morning peak passenger flow of the first subway station, including: performing a passenger flow core feature aggregation analysis based on time series features on the historical data of the morning peak passenger flow of the first subway station to obtain the predicted morning peak passenger flow of the first subway station; On the cloud platform, based on a comparison between the morning peak passenger flow of the first subway station and the predicted morning peak passenger flow of the first subway station, it is determined whether to generate an abnormal passenger flow alarm prompt.

2. The method for analyzing citizen card consumption data based on a cloud platform according to claim 1 is characterized in that: Performing a passenger flow core feature aggregation analysis based on time series features on the historical data of the morning peak passenger flow of the first subway station to obtain the predicted morning peak passenger flow of the first subway station, including: Performing data sorting and periodic sequence feature extraction on historical data of morning peak passenger flow at the first subway station to obtain multiple time subsequences of morning peak passenger flow; Performing passenger flow time series feature extraction on each of the multiple morning peak passenger flow time subsequences to obtain multiple morning peak passenger flow time series feature encoding vectors; Performing a passenger flow time series multi-period feature significant aggregation analysis on the multiple morning peak passenger flow time series feature coding vectors to obtain a morning peak passenger flow multi-period significant aggregation coding vector; Based on the multi-time series period significant aggregation coding vector of the morning peak passenger flow, the predicted morning peak passenger flow of the first subway station is obtained.

3. The method for analyzing citizen card consumption data based on a cloud platform according to claim 2, characterized in that: The historical data of the morning peak passenger flow of the first subway station is sorted and periodic sequence features are extracted to obtain multiple time subsequences of the morning peak passenger flow, including: The historical data of the morning peak passenger flow of the first subway station is sorted according to the time dimension to obtain a time series of the morning peak passenger flow; The time series of the morning peak passenger flow is processed using a singular spectrum decomposition algorithm to extract the periodic characteristics of the time series of the morning peak passenger flow to obtain the multiple time subsequences of the morning peak passenger flow.

4. The method for analyzing citizen card consumption data based on a cloud platform according to claim 3 is characterized in that: Performing passenger flow time series feature extraction on each time subsequence of the multiple morning peak passenger flows to obtain multiple morning peak passenger flow time series feature coding vectors, including: using a Bi-GRU-based feature extractor to process each time subsequence of the multiple morning peak passenger flows to obtain the multiple morning peak passenger flow time series feature coding vectors.

5. The method for analyzing citizen card consumption data based on a cloud platform according to claim 4 is characterized in that: Performing a passenger flow time series multi-period feature significant aggregation analysis on the multiple morning peak passenger flow time series feature coding vectors to obtain a morning peak passenger flow multi-period significant aggregation coding vector, including: Performing passenger flow information core coarse-grained aggregation on the multiple morning peak passenger flow time series feature coding vectors to obtain a morning peak passenger flow time series coarse-grained aggregation coding vector; Based on the morning peak passenger flow time series coarse-grained aggregated coding vector, gated explicit fine-grained compensation aggregation is performed on the multiple morning peak passenger flow time series feature coding vectors to obtain a passenger flow time series fine-grained compensation aggregated coding vector; The morning peak passenger flow time series coarse-grained aggregation coding vector and the passenger flow time series fine-grained compensation aggregation coding vector are interactively analyzed to obtain the morning peak passenger flow multi-time series period significant aggregation coding vector.

6. The method for analyzing citizen card consumption data based on a cloud platform according to claim 5 is characterized in that: Based on the morning peak passenger flow time series coarse-grained aggregated coding vector, gated explicit fine-grained compensation aggregation is performed on the multiple morning peak passenger flow time series feature coding vectors to obtain a passenger flow time series fine-grained compensation aggregated coding vector, including: Calculating a core convergence compensation factor of each morning peak passenger flow time series feature coding vector in the multiple morning peak passenger flow time series feature coding vectors relative to the morning peak passenger flow time series coarse-grained convergence coding vector to obtain multiple passenger flow time series core convergence compensation factors; Performing gated explicit compensation on the plurality of passenger flow timing core convergence compensation factors to obtain a plurality of passenger flow timing core convergence compensation weight factors; The multiple passenger flow time series core aggregation compensation weight factors, the morning peak passenger flow time series coarse-grained aggregation coding vectors and the multiple morning peak passenger flow time series feature coding vectors are subjected to passenger flow time series fine-grained dynamic compensation aggregation to obtain the passenger flow time series fine-grained compensation aggregation coding vector.

7. The method for analyzing citizen card consumption data based on a cloud platform according to claim 6 is characterized in that: Calculating the core convergence compensation factor of each morning peak passenger flow time series feature coding vector in the multiple morning peak passenger flow time series feature coding vectors relative to the morning peak passenger flow time series coarse-grained convergence coding vector to obtain multiple passenger flow time series core convergence compensation factors, including: Calculating a morning peak passenger flow time series difference coding vector between the morning peak passenger flow time series feature coding vector and the morning peak passenger flow time series coarse-grained aggregation coding vector; The morning peak passenger flow time series differential coding vector is multiplied by the passenger flow time series weight matrix, and then added to the passenger flow time series offset value position by position to obtain the morning peak passenger flow time series compensation vector; Multiplying the morning peak passenger flow time series compensation vector by the scoring weight vector to obtain a passenger flow time series kernel convergence compensation factor corresponding to the morning peak passenger flow time series feature coding vector; In response to the fact that the second norm of the enhanced morning peak passenger flow time series feature coding vector is less than the second norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector, a logarithmic function value obtained by adding a constant one to the ratio of the second norm of the enhanced morning peak passenger flow time series feature coding vector to the second norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector is taken as the passenger flow time series offset value; In response to the fact that the binary norm of the enhanced morning peak passenger flow time series feature coding vector is greater than or equal to the binary norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector, the ratio of the binary norm of the enhanced morning peak passenger flow time series feature coding vector and the binary norm of the enhanced morning peak passenger flow time series coarse-grained aggregation coding vector is used as the passenger flow time series bias value.

8. The method for analyzing citizen card consumption data based on a cloud platform according to claim 7 is characterized in that: Based on the multi-time-series-period-significant-aggregation coding vector of the morning peak passenger flow, the predicted morning peak passenger flow of the first subway station is obtained, including: inputting the multi-time-series-period-significant-aggregation coding vector of the morning peak passenger flow into a morning peak passenger flow prediction module based on the RNN model to obtain the predicted morning peak passenger flow of the first subway station.

9. The method for analyzing citizen card consumption data based on a cloud platform according to claim 8, characterized in that: Determining whether to generate an abnormal passenger flow alarm based on a comparison between the morning peak passenger flow at the first subway station and the predicted morning peak passenger flow at the first subway station includes: Calculating a ratio between the morning peak passenger flow of the first subway station and the predicted morning peak passenger flow of the first subway station as a passenger flow deviation rate; Based on the comparison between the passenger flow deviation rate and a preset threshold, it is determined whether to generate the passenger flow abnormality alarm prompt.

10. A citizen card consumption data analysis system based on a cloud platform, characterized in that: include: The subway station card swiping data acquisition module is used to obtain the subway station card swiping data of the first subway station during the morning rush hour; A subway station card swiping data uploading module, used to upload the subway station card swiping data of the first subway station to the cloud platform; A morning peak passenger flow statistics module is used to count the morning peak passenger flow of the first subway station based on the subway card swiping data of the first subway station on the cloud platform; A passenger flow historical data extraction module extracts the historical data of the morning peak passenger flow of the first subway station from the backend database on the cloud platform; a morning peak passenger flow prediction module, configured to predict the morning peak passenger flow of the day on the cloud platform based on the historical data of the morning peak passenger flow of the first subway station to obtain the predicted morning peak passenger flow of the first subway station, wherein the morning peak passenger flow prediction module comprises: performing a passenger flow core feature aggregation analysis based on time series features on the historical data of the morning peak passenger flow of the first subway station to obtain the predicted morning peak passenger flow of the first subway station; The passenger flow abnormality alarm module is used to determine whether to generate a passenger flow abnormality alarm prompt on the cloud platform based on the comparison between the morning peak passenger flow of the first subway station and the predicted morning peak passenger flow of the first subway station.

Images