Travel big data analysis method and system based on traffic card swiping information

Abstract

The application relates to the technical field of traffic control, in particular to a travel big data analysis method and system based on traffic card swiping information, which solves the technical problem that the influence of uncertainty between nodes on a travel chain structure leads to a low reduction degree of the travel chain construction in the prior art. The method comprises the following steps: generating a card swiping event node sequence; listing a plurality of candidate carrying structures, and determining a chain structure refinement requirement degree according to the differences between the chain structures corresponding to the plurality of candidate carrying structures; supplementing travel process information according to the chain structure refinement requirement degree, obtaining a refined structure, and determining the carrying credibility of each candidate carrying structure according to the structure conflict degree between each candidate carrying structure and the refined structure; taking the candidate carrying structure with the highest carrying credibility as the optimal carrying structure, and generating a user travel chain.

Description

Technical Field

[0001] This invention relates to the field of traffic control technology, specifically to a method and system for analyzing travel big data based on traffic card swiping information. Background Technology

[0002] In public transportation systems, card swiping behaviors such as using bus cards, subway cards, and QR codes generate a large number of transaction records. These records typically come from subway turnstiles, onboard terminals, transfer channel equipment, and mobile payment settlement platforms, and include fields such as user anonymity identifier, swipe time, device, station, vehicle, route, and swipe type. Because card swipe records have strong location anchoring, as long as the record actually occurs, it can be confirmed that the user was at the corresponding location at that time, thus becoming an important data source for analyzing user travel behavior.

[0003] However, card swipe records can only fix discrete nodes on the travel chain and cannot determine the actual traffic process between two nodes: in the bus scenario, there is often only card swiping when boarding but not when alighting; in the subway scenario, although entry and exit records exist, transfer channels, stops at hubs, and re-entry after exiting the station can lead to different interpretations; QR code payment may also form abnormal nodes due to offline re-upload, duplicate charges, or device time deviations; adjacent card swipe nodes may be a normal transfer, or they may be a missing alighting node, a short stop, the start of an independent trip, or an abnormal card swipe. These uncertainties make it difficult to accurately reflect the real travel chain by directly splicing records according to time or forcibly reconstructing the complete trajectory.

[0004] In existing technologies, the common practice is to directly splice adjacent records in chronological order to construct the travel chain, or to restore the complete trajectory by adding a large number of candidate paths. This approach cannot distinguish whether the uncertain processes between nodes have actually changed the boarding and alighting stations, transfer points, and the location where the travel chain is broken, which may lead to excessive splitting, incorrect merging, or trajectory restoration during the travel chain construction process. Summary of the Invention

[0005] To address the technical problem of low accuracy in travel chain reconstruction caused by uncertainties between nodes in existing technologies, this invention aims to provide a travel big data analysis method and system based on traffic card swiping information. The specific technical solution adopted is as follows: Firstly, a method for analyzing travel big data based on traffic card swiping information is provided, including: generating a sequence of card swiping event nodes based on traffic card swiping records; listing multiple candidate connection structures between adjacent card swiping event nodes, and determining the refinement requirement of the chain structure between adjacent card swiping event nodes based on the differences between the chain structures corresponding to the multiple candidate connection structures; supplementing traffic process information between adjacent card swiping event nodes based on the refinement requirement of the chain structure to obtain a refined structure, and determining the connection credibility of each candidate connection structure based on the degree of structural conflict between each candidate connection structure and the refined structure; and taking the candidate connection structure with the highest connection credibility as the optimal connection structure between adjacent card swiping event nodes to generate a user travel chain.

[0006] Based on the above technical solution, in the travel big data analysis method based on traffic card swiping information provided by this invention, the traffic process between adjacent card swiping event nodes is abstracted into candidate receiving structures, and further mapped into chain structures for difference analysis to obtain the chain structure refinement requirement. This allows the supplementation of traffic process information to be implemented in a targeted manner based on the degree of difference at the chain structure level. On this basis, by comparing the degree of structural conflict between each candidate receiving structure and the refined structure obtained after supplementation, the candidate receiving structure with the highest receiving credibility is selected as the optimal receiving structure to generate the user travel chain. Thus, the construction of the travel chain is improved from a simple time sequence splicing to a refined processing based on chain structure difference analysis and structural conflict comparison, so that the generation process of the travel chain can reflect the divergence influence of different candidate receiving structures at the chain structure level.

[0007] In conjunction with the first aspect above, in one possible implementation, the method of listing multiple candidate succession structures between adjacent card swiping event nodes specifically includes: generating candidate succession structures of at least one type, including transfer, missing boarding / alighting, short stop, start of independent travel, and abnormal card swiping, based on the time interval, spatial location relationship, route connection relationship, and mode of transportation conversion relationship between adjacent card swiping event nodes.

[0008] In conjunction with the first aspect above, in one possible implementation, the method for determining the chain structure refinement requirement between adjacent card-swiping event nodes based on the differences between the chain structures corresponding to multiple candidate receiving structures specifically includes: extracting at least one chain node position from each candidate receiving structure, including boarding / alighting stations, transfer points, and disconnection points of the travel chain, and mapping them to the corresponding chain structure; counting the number of duplicate chain structures corresponding to multiple candidate receiving structures to determine the node receiving divergence degree between adjacent card-swiping event nodes; counting the differences in chain node positions between the chain structures corresponding to multiple candidate receiving structures; and combining the node receiving divergence degree with the differences in chain node positions to determine the chain structure refinement requirement degree.

[0009] In conjunction with the first aspect above, in one possible implementation, the method of supplementing traffic process information between adjacent card-swiping event nodes based on the chain structure refinement demand degree to obtain a refined structure specifically includes: if the chain structure refinement demand degree is greater than a preset demand threshold, then calling station topology data, line station sequence data, transfer channel data, or train schedule information to supplement at least one chain node position among the boarding / alighting stations, transfer points, and disconnection points of the travel chain between adjacent card-swiping event nodes to form a refined structure; if the chain structure refinement demand degree is less than or equal to the preset demand threshold, then any candidate receiving structure is used as the refined structure.

[0010] In conjunction with the first aspect mentioned above, in one possible implementation, the method for determining the acceptance credibility of each candidate acceptance structure based on the degree of structural conflict between each candidate acceptance structure and the refinement structure specifically includes: comparing the chain node positions of the refinement structure item by item according to the type of each candidate acceptance structure, counting the number of conflicts at the chain node positions, and determining the degree of structural conflict; performing a negative correlation mapping based on the degree of structural conflict to obtain the acceptance credibility of each candidate acceptance structure.

[0011] In conjunction with the first aspect above, in one possible implementation, the method for counting the number of conflicts at chain node locations specifically includes: if the candidate receiving structure is a transfer type, and the time interval between adjacent chain nodes in the refined structure is greater than a preset transfer time threshold or the spatial distance between adjacent chain nodes is greater than a preset walking distance threshold, then the number of conflicts is accumulated; if the candidate receiving structure is a missing boarding / alighting type, and the positional span between the supplemented boarding / alighting station and the subsequent card swiping event node in the refined structure is greater than a preset connection span threshold, then the number of conflicts is accumulated; if the candidate receiving structure is a short stay type, and the time interval between adjacent chain nodes in the refined structure is greater than a preset short stay time threshold or the spatial locations of adjacent chain nodes belong to different stations, then the number of conflicts is accumulated; if the candidate receiving structure is an independent trip start type, and there is a preset traffic continuity relationship between the preceding and following chain nodes in the refined structure, then the number of conflicts is accumulated; if the candidate receiving structure is an abnormal card swiping type, and the spatiotemporal correlation characteristics between the chain nodes and the card swiping event node in the refined structure meet preset correlation conditions, then the number of conflicts is accumulated.

[0012] In conjunction with the first aspect mentioned above, in one possible implementation, the method of generating a user travel chain by using the candidate connection structure with the highest connection credibility as the optimal connection structure between adjacent card swipe event nodes specifically includes: if the optimal connection structure is a transfer type, merging adjacent card swipe event nodes into the same travel chain and recording the transfer point; if the optimal connection structure is a missing boarding / alighting type, inserting the missing boarding / alighting station as a new node into the travel chain; if the optimal connection structure is a short stop type, merging adjacent card swipe event nodes into the same travel chain; if the optimal connection structure is an independent travel start type, using the next card swipe event node as the starting point of a new travel chain; if the optimal connection structure is an abnormal card swipe type, removing the corresponding card swipe event node.

[0013] In conjunction with the first aspect above, in one possible implementation, the method for generating a card swipe event node sequence based on traffic card swipe records specifically includes: obtaining original card swipe records; processing each original card swipe record to remove records lacking user identifiers or swipe times, and filling in records lacking location information based on vehicle location and route station sequence; converting each processed original card swipe record into a card swipe event node, grouping them according to user identifiers, and arranging them within each group according to swipe time to obtain a card swipe event node sequence.

[0014] In conjunction with the first aspect mentioned above, in one possible implementation, after generating the user travel chain, the method further includes: arranging the travel chains generated by the same user within the same operating cycle in chronological order to form a user travel chain sequence; and extracting the starting point, ending point, transfer point, mode of transport combination, and time window of each travel chain based on the user travel chain sequence, and performing statistics on the origin and destination flow of passengers, statistics on transfer relationships, identification of commuting patterns, or analysis of regional passenger flow changes.

[0015] Secondly, a travel big data analysis system based on traffic card swiping information is provided, including: a node sequence generation module, used to generate a sequence of card swiping event nodes based on traffic card swiping records; a detailed requirement determination module, used to list multiple candidate connection structures between adjacent card swiping event nodes, and determine the detailed requirement degree of the chain structure between adjacent card swiping event nodes based on the differences between the chain structures corresponding to multiple candidate connection structures; a credibility analysis module, used to supplement traffic process information between adjacent card swiping event nodes based on the detailed requirement degree of the chain structure, to obtain a detailed structure, and determine the connection credibility of each candidate connection structure based on the degree of structural conflict between each candidate connection structure and the detailed structure; and a travel chain generation module, used to select the candidate connection structure with the highest connection credibility as the optimal connection structure between adjacent card swiping event nodes, and generate a user travel chain.

[0016] The present invention has the following beneficial effects: By abstracting the traffic process between adjacent card-swiping event nodes into candidate receiving structures and further mapping them into chain structures for difference analysis to obtain the chain structure refinement requirement, the supplementation of traffic process information can be implemented in a targeted manner based on the degree of difference at the chain structure level. On this basis, by comparing the degree of structural conflict between each candidate receiving structure and the refined structure obtained after supplementation, the candidate receiving structure with the highest receiving credibility is selected as the optimal receiving structure to generate the user travel chain. Thus, the construction of the travel chain is improved from a simple time sequence splicing to a refined processing based on chain structure difference analysis and structural conflict comparison, so that the travel chain generation process can reflect the divergence impact of different candidate receiving structures at the chain structure level. Attached Figure Description

[0017] Figure 1 A system structure diagram of a travel big data analysis system based on traffic card swiping information provided in one embodiment of the present invention; Figure 2 The present invention provides a flowchart of a method for analyzing travel big data based on traffic card swiping information, which is an embodiment of the present invention. Detailed Implementation

[0018] The following description, in conjunction with the accompanying drawings, details the specific scheme of a travel big data analysis method and system based on traffic card swiping information provided by the present invention.

[0019] Please see Figure 1 The diagram illustrates a system architecture of a travel big data analysis system based on traffic card swiping information according to an embodiment of the present invention. The travel big data analysis system based on traffic card swiping information includes: a node sequence generation module 1, a detailed demand determination module 2, a credibility analysis module 3, and a travel chain generation module 4.

[0020] The node sequence generation module 1 serves as the system's data entry point, typically deployed on a data server or edge computing node. It acquires raw card-swiping records from subway turnstiles, bus terminals, QR code payment platforms, and transfer channel equipment in real-time or offline via interfaces. This module first validates each record, removing invalid records lacking user identifiers or swipe times. For records missing station location information, spatial location is filled in based on vehicle positioning data and the route station sequence table. Subsequently, each cleaned record is standardized into a card-swiping event node, with each node containing at least a user anonymity identifier, swipe time, traffic location code, route code, mode of transport, and card type. Finally, this module groups the nodes according to user identifiers, arranging them in ascending order of swipe time within each group to form a card-swiping event node sequence. The generated sequence is then passed to the detailed requirements determination module 2 as the foundational data unit for subsequent analysis.

[0021] The detailed requirement determination module 2 receives the card-swiping event node sequence from the node sequence generation module 1 and is responsible for systematically analyzing the uncertainty between each pair of adjacent nodes. This module first lists multiple candidate connection structures based on the time interval, spatial location, route connection, and mode of transport conversion relationships between adjacent nodes, including types such as transfers, missing boarding / alighting, short stops, independent travel starts, and abnormal card swiping. For each candidate connection structure, the module extracts its impact on the travel chain structure, specifically manifested as the location of chain nodes such as boarding / alighting stations, transfer points, and travel chain break points, and maps them to the corresponding chain structure. Subsequently, the module counts the number of duplicate chain structures corresponding to all candidate connection structures to determine the node connection divergence. It also counts the differences in chain node locations between different chain structures. This module combines the node connection divergence with the differences in chain node locations to calculate the chain structure detailed requirement degree. This requirement degree quantifies whether additional traffic process information is needed between current adjacent nodes to distinguish the actual impact of different candidate structures on the final travel chain. The calculated detailed requirement degree, along with the original node sequence and candidate connection structures, is passed to the credibility analysis module 3.

[0022] The credibility analysis module 3 determines the refinement requirement of the chain structure output by module 2 based on the refinement needs, and decides whether to perform limited refinement operations. If the refinement requirement is greater than the preset threshold, the module calls auxiliary data such as the station topology database, line station sequence table, transfer passage geographical information, and bus / railway timetable to supplement the necessary chain node positions between adjacent card-swiping event nodes. For example, it infers missing alighting stations, determines connection points in transfer passages, or marks the location of travel chain breaks, thereby forming a refined structure. If the refinement requirement is not greater than the threshold, no supplementary information is needed, and the module directly selects any candidate receiving structure (usually the one closest in time or most reasonable in space) as the refined structure. After obtaining the refined structure, the module compares the chain node positions in each candidate receiving structure with those in the refined structure item by item, and counts the number of conflicts between the two in terms of alighting / alighting stations, transfer relationships, time intervals, spatial distances, and traffic continuation logic, using this as the degree of structural conflict. Finally, the module performs a negative correlation mapping on the degree of structural conflict (e.g., using the reciprocal of the number of conflicts or an exponential decay function) to calculate the acceptance credibility of each candidate receiving structure. The fewer the conflicts, the higher the credibility. This module marks the candidate receiving structure with the highest credibility as the optimal receiving structure, and outputs the optimal receiving structure along with the original node sequence to the outgoing chain generation module 4.

[0023] The travel chain generation module 4 receives the optimal connection structure and corresponding adjacent card-swiping event nodes output by the credibility analysis module 3. This module performs different travel chain construction operations based on the type of the optimal connection structure: for transfer types, it merges two adjacent card-swiping event nodes into the same travel chain and records the transfer point. For missing boarding / alighting points, it inserts the missing boarding / alighting points as new nodes into the travel chain. For short-stay types, it also merges adjacent nodes into the same travel chain (without additional transfer point marking). For independent travel start types, it splits at the next card-swiping event node, using the next node as the starting point of a new travel chain. For abnormal card-swiping types, it removes the corresponding nodes to prevent abnormal data from interfering with the integrity of the travel chain. This module processes all adjacent node pairs sequentially, ultimately generating a complete and stable travel chain for each user. The generated travel chain can be further used for upper-level applications such as passenger flow origin-destination direction statistics, transfer relationship analysis, commuting pattern identification, and regional passenger flow change prediction.

[0024] The four modules can be deployed on the same high-performance server or distributed across the data acquisition layer, computing layer, and application layer. Module 1, the node sequence generation module, is typically located close to the data source (e.g., a data access gateway in a transportation data center). Modules 2, the detailed requirements determination module, and 3, the credibility analysis module, run on a computing cluster or cloud computing platform. Module 4, the travel chain generation module, outputs analysis results to the business application. Intermediate results are exchanged between modules via standardized data interfaces, supporting both batch offline processing and real-time streaming processing modes.

[0025] Please see Figure 2 The diagram illustrates a flowchart of a travel big data analysis method based on traffic card swipe information according to an embodiment of the present invention. This travel big data analysis method based on traffic card swipe information includes: S1. Generate a sequence of card swipe event nodes based on the traffic card swipe records.

[0026] The original transportation card swipe records originate from multiple heterogeneous systems such as subway turnstiles, bus terminals, and QR code payment platforms. Their field definitions, data formats, and quality vary considerably. Furthermore, these records only determine a user's location at a specific time and cannot be directly used to analyze the continuity between adjacent swipe actions. By uniformly transforming the original records into standardized card swipe event nodes, the heterogeneity of multi-source data can be eliminated. Each node serves as a spatial anchor point with a definite time and location, representing a key, evidence-supported position in the user's travel chain. The resulting sequence of card swipe event nodes, arranged according to user identifier and swipe time, preserves the complete node order for the same user on the timeline.

[0027] In some implementations, raw card-swipe records are obtained and processed. If a record lacks a user identifier or swipe time, it cannot be associated with a specific user or its time sequence cannot be determined, and is therefore discarded. If a record lacks location information but contains vehicle location data and route station sequence, it is supplemented with the corresponding traffic location code based on the vehicle location and route station sequence. Furthermore, for multiple duplicate charge records generated by the same device within a preset duplicate detection time window (e.g., five seconds), only the earliest record is retained, and the remaining records are marked as duplicate anomalies and excluded from the node sequence generation process.

[0028] Each processed original card swipe record is transformed into a card swipe event node. A node represents a user's spatiotemporal anchor point supported by evidence. Its data structure includes at least the user identifier, swipe time, transportation location code, line code, mode of transportation, and card swipe type. For example, a card swipe event node can be represented as follows: user identifier "USER_12345", swipe time "June 9, 2026, 08:35:22", transportation location code "STN_1001" (corresponding to a certain subway station), line code "LINE_2" (corresponding to subway line 2), mode of transportation "subway", and card swipe type "enter station". In another example, the node generated by the bus terminal can be represented as follows: user identifier is "USER_67890", card swipe time is "08:42:05 on June 9, 2026", traffic location code is "BUS_0523" (corresponding to the code after latitude and longitude conversion of a bus stop or vehicle location), route code is "BUS_202", mode of transportation is "bus", and card swipe type is "boarding".

[0029] Simultaneously, each record retains a source marker (to distinguish between turnstiles, vehicle terminals, QR code platforms, or transfer channel equipment) and a quality marker (to indicate whether there are any retransmissions, duplicates, missing stations, or time offsets). These markers are used to determine whether to downgrade or remove nodes when there are conflicts in subsequent candidate connection structures. Subsequently, the nodes are grouped according to user identifiers, and within each group, they are arranged by card swipe time to obtain a sequence of card swipe event nodes. Between two adjacent nodes in the sequence, there is a corresponding interval to be interpreted, used to express the state where the connection relationship between the preceding and following nodes is not yet determined.

[0030] S2. List multiple candidate connection structures between adjacent card swipe event nodes, and determine the degree of refinement requirement of the chain structure between adjacent card swipe event nodes based on the differences between the chain structures corresponding to the multiple candidate connection structures.

[0031] The traffic process between two adjacent card-swipe event nodes involves various uncertainties, which may manifest as different types of connection relationships (e.g., whether the user transfers, whether there is a missing alighting record, whether the user continues their journey after a short stop, whether a new independent journey begins, or whether the card-swipe record itself is abnormal). Each connection relationship, if adopted, will have different impacts on the final structure of the user's travel chain (including the origin and destination positions of the travel chain, whether there are transfer points, and whether it needs to be divided into two independent travel chains). However, not all uncertainties will lead to substantial differences in the travel chain structure; for example, two different connection relationships may ultimately point to the exact same origin, destination, and transfer point. Therefore, it is necessary to first list all logically possible candidate connection structures, and then compare the degree of difference between the travel chain structures corresponding to these candidate structures to determine whether it is necessary to further refine the interval. If the degree of difference is high, it indicates that the uncertainty of this interval will significantly affect the travel chain structure, requiring further information to distinguish different candidates; if the degree of difference is low, no refinement is needed, and the connection relationship can be directly selected according to the default rules.

[0032] In some implementations, candidate succession structures are generated based on the time interval, spatial location relationship, route connection relationship, and mode of transportation conversion relationship between adjacent card swipe event nodes. These structures include at least one type of candidate succession structure, such as transfer, missing boarding / alighting, short stop, start of independent travel, and abnormal card swipe. Specifically, if the time interval is less than a preset transfer time threshold (e.g., 30 minutes based on urban traffic characteristics, which can be relaxed to 45 minutes during peak commuting hours) and the spatial distance is less than a preset walking distance threshold (e.g., 500 to 800 meters, set based on the walking accessibility between stations), a "transfer" type is generated; if the previous node is a bus boarding and the next node is a subway entrance or bus boarding, but there is no alighting record in between, a "missing boarding / alighting" type is generated; if the time interval is short (e.g., less than 10 minutes) and the spatial locations of the preceding and following nodes are the same or adjacent, a "short stop" type is generated; if the time interval or spatial distance between the preceding and following nodes far exceeds the normal transfer range (e.g., the interval exceeds 2 hours or the distance exceeds 10 kilometers), a "start of independent travel" type is generated; if the quality marker carried by the node indicates abnormalities such as retransmission, duplication, or time offset, an "abnormal card swipe" type is generated. The above candidate receiving structures constitute the candidate set for this interval.

[0033] From each candidate connection structure, at least one of the following chain node positions is extracted: boarding / alighting station, transfer point, and chain break point. This is mapped to the corresponding chain structure. The chain structure is an abstract description of the key node positions of the travel chain after the candidate connection relationship is adopted, and it includes at least one of the following: boarding / alighting station, transfer point, and chain break point. For example, for the "transfer" type, the chain structure will mark the location where the transfer occurs (usually the location of the next card swipe event node); for the "missing boarding / alighting" type, the chain structure will mark a missing station placeholder. This placeholder is only used to indicate that there is a station to be filled at the chain node location and does not participate in the specific numerical calculation of spatial location and time difference when calculating the differences in chain node locations; it only participates in the existence comparison; for the "independent trip start" type, the chain structure will mark a chain break point between two nodes, indicating that the latter node is the starting point of the new trip.

[0034] The deduplication count of chain structures corresponding to multiple candidate receiving structures is calculated. The deduplication count refers to the number of different categories remaining after treating identical chain structures as the same type. For example, if the chain structures mapped to the three candidate structures "transfer," "short stop," and "missing boarding / alighting" are completely identical in terms of boarding / alighting stations, transfer points, and break points, the deduplication count is 1. If only the chain structure corresponding to "start of independent travel" differs from the others (e.g., it has an extra break point marker), the deduplication count is 2. Then, the node receiving divergence degree between adjacent card-swiping event nodes is determined based on the relationship between the deduplication count and the total number of candidate receiving structures. The chain structure result mapped to each candidate structure is the smallest indivisible unit for judging divergence. The impact of the deduplication count on the divergence degree is balanced by the total number of candidate receiving structures. When there are many candidates but few chain structure types, the divergence degree is low, reflecting the consensus effect of most candidates pointing to the same structure; when there are few candidates but the same number of chain structure types, the divergence degree is high, reflecting the conflict effect of a few candidates each holding their own opinion. An exemplary calculation method is as follows: In the formula, G represents the node acceptance divergence degree, indicating the degree of difference in the chain structure result caused by different candidate acceptance structures between adjacent card swipe event nodes, with a value range of [0, 1]. S represents the deduplication quantity, indicating the number of elements in the set obtained after deduplicating the chain structure after mapping all candidate acceptance structures to the corresponding chain structure. C represents the total number of candidate acceptance structures.

[0035] When the deduplication count is 1, it means that all candidate structures result in identical chain structures after mapping. In this case, there is no structural divergence, and the divergence degree should be 0. Therefore, the molecule adopts... When the number of duplicates removed is greater than 1, the degree of divergence is positive, and the larger the number of duplicates removed, the larger the numerator and the higher the degree of divergence. This linear relationship is intuitive and easy to understand. The corresponding denominator and numerator use the same offset, i.e. When the number of candidates is 1, there is no divergence comparison. Using the maximum value function max to make the denominator 1 makes the divergence degree 0, which is logically reasonable.

[0036] Simultaneously, the differences in chain node positions between chain structures corresponding to multiple candidate receiving structures are statistically analyzed. Specifically, for the chain structures mapped by different candidate receiving structures, the positional attributes of chain nodes are extracted item by item and compared pairwise to quantify the difference values. Chain node positional attributes include spatial and temporal attributes. Spatial location attributes are the traffic location codes or geographic coordinates corresponding to the chain nodes, and temporal attributes are the time intervals between adjacent chain nodes. For spatial location differences, the spatial distance between corresponding chain nodes in the chain structures corresponding to two candidate structures is calculated and normalized by dividing by a preset maximum spatial distance benchmark (e.g., 10 kilometers). For temporal attribute differences, the time interval difference between corresponding chain nodes in the chain structures corresponding to two candidate structures is calculated and normalized by dividing by a preset maximum time difference benchmark (e.g., 2 hours). The normalized spatial location differences and temporal attribute differences are then weighted and summed to obtain the chain node positional difference value between each pair of candidate structures. The weighting coefficient for spatial location differences is 0.6, and the weighting coefficient for temporal attribute differences is 0.4. These weighting coefficients are set based on practical experience that the requirement for spatial anchor point positioning accuracy is higher than that for temporal accuracy in traffic travel analysis. The maximum value of the chain node position difference between all candidate pairs is taken as the chain node position difference between adjacent card swipe event nodes, and the value range is [0, 1].

[0037] Finally, the refinement requirement for the chain structure is determined by combining the node acceptance divergence degree and the chain node position difference. An exemplary calculation method is to multiply the node acceptance divergence degree by the chain node position difference to obtain the refinement requirement. The node acceptance divergence degree reflects the degree of divergence of candidate structures at the qualitative level of "whether they lead to different chain structures," but it does not consider the magnitude of the specific differences between different chain structures. The chain node position difference quantitatively describes the degree of difference, but it cannot distinguish whether the difference stems from genuine structural divergence or spurious divergence caused by data noise or anomalies. Multiplying the two allows the refinement requirement to reflect both the divergence and the severity of the divergence. When the node acceptance divergence is zero, the refinement requirement is zero, indicating that all candidate acceptance structures point to the exact same chain structure, and no additional traffic process information is needed. When the chain node position difference is zero, the refinement requirement is zero, indicating that although there is divergence in the qualitative classification of candidate structures, there is no substantial difference in spatial location and temporal attributes between different structures, and the actual impact on the travel chain structure can be ignored. Only when both the node acceptance divergence and the chain node position are positive is the refinement requirement positive, triggering subsequent refinement operations.

[0038] S3. Based on the chain structure refinement requirement, supplement the traffic process information between adjacent card swiping event nodes to obtain the refined structure, and determine the acceptance credibility of each candidate acceptance structure based on the degree of structural conflict between each candidate acceptance structure and the refined structure.

[0039] The detailed chain structure requirement quantifies whether additional traffic process information needs to be supplemented to differentiate the impact of different candidate receiving structures on the travel chain structure. A high detailed requirement indicates significant discrepancies in the chain node positions corresponding to different candidate structures. Auxiliary data must be used to supplement missing chain node information (such as disembarkation stations, transfer connection points, or travel chain breakpoints) to obtain a more realistic detailed structure, serving as a benchmark for evaluating the credibility of each candidate structure. A low detailed requirement indicates negligible differences in the impact of different candidate structures on the travel chain structure. No auxiliary data is needed, and any candidate structure (e.g., the first candidate appearing in chronological order, or a candidate selected according to a preset priority) can be directly used as the detailed structure. Subsequently, by comparing each candidate receiving structure with the detailed structure item by item and counting the number of conflicts at chain node positions, the degree of inconsistency between each candidate structure and the detailed result can be quantified. Fewer conflicts indicate a better match between the candidate structure and the supplemented traffic process information, resulting in higher credibility; more conflicts indicate lower credibility. Finally, the degree of conflict is converted into the credibility of acceptance through negative correlation mapping, providing a quantitative basis for selecting the optimal acceptance structure.

[0040] In some implementations, a threshold value is preset. Specifically, a boundary value that can balance refinement accuracy and computational efficiency can be selected based on historical data or experience. For example, the threshold value can be set to 0.6.

[0041] If the required refinement of the chain structure exceeds the preset threshold, it is considered that the uncertainty in this interval has a significant impact on the travel chain structure, and limited refinement needs to be performed. At this time, pre-stored auxiliary data is invoked, including station topology data (describing the spatial connectivity and distance between stations), line station sequence data (the station order of each bus or subway line), transfer channel data (walking paths and times between different lines within the hub), or train information (vehicle departure and arrival times). At least one chain node position among the boarding / alighting stations, transfer points, and disconnection points of the travel chain is supplemented between adjacent card-swiping event nodes to form a refined structure. During this process, according to the specific operation category of supplementing by calling auxiliary data, the refined structure is synchronously assigned a corresponding type label (such as transfer type, boarding / alighting missing type, etc.) and written into its data structure. For example, if the previous node is a bus boarding point and the next node is a subway entry point, but there is no bus alighting record in between, then the most likely alighting station is inferred as the supplementary chain node based on the bus route's stop sequence and the location of the next subway station. If there may be a transfer between the two nodes, the transfer point location is supplemented based on the transfer channel data. If it is determined that the travel chain needs to be broken, the break point is marked between the two nodes. Through the above limited refinement, a refined structure containing supplementary information is formed.

[0042] If the chain structure refinement requirement is less than or equal to the preset requirement threshold, it indicates that there is no substantial difference between the chain structures corresponding to each candidate receiving structure, and there is no disagreement affecting the final travel chain structure. In this case, there is no need to call auxiliary data or perform conflict comparison between candidate receiving structures and refined structures. Any candidate receiving structure is used as a refined structure. Preferably, the optimal receiving structure is selected from the candidate receiving structures according to the preset priority rules. The preset priority rules are as follows: "transfer" type is selected first, followed by "missing boarding / alighting" type, then "short stop" type, and then "independent travel start" type. The "abnormal card swipe" type is only selected when the credibility is significantly higher than other types (e.g., the credibility difference is greater than 0.3). Otherwise, other types are selected first to avoid excessive data rejection. If there are multiple types with the same priority in the candidate structures, the candidate receiving structure with the smallest spatial distance from the previous card swipe event node can be selected.

[0043] When the chain structure refinement requirement exceeds the preset threshold, after obtaining the refined structure, the chain node positions of the refined structure are compared item by item according to the type of each candidate receiving structure. The number of conflicts at the chain node positions is counted to determine the degree of structural conflict. First, by reading the type tags written above, it is confirmed whether the refined structure contains the type corresponding to the candidate receiving structure. If the refined structure does not contain the corresponding type, it is determined that the structure comparison is missing, and the number of conflicts is directly set to the historical maximum value (if the current system has no historical data accumulation, the default value is assigned to the preset maximum penalty constant), and the comparison of the refined structure corresponding to the candidate receiving structure is stopped. If the refined structure contains the corresponding type, the specific comparison rules are as follows: If the candidate transfer structure is a transfer type, the system checks whether the time interval between adjacent chain nodes in the refined structure exceeds a preset transfer time threshold, or whether the spatial distance between adjacent chain nodes exceeds a preset walking distance threshold. The transfer time threshold is set based on urban traffic characteristics; for example, it is set at 30 minutes in ordinary cities, can be relaxed to 45 minutes during peak hours, and shortened to 20 minutes at night. The walking distance threshold is determined based on the walkability between stations, typically between 500 and 800 meters; for example, a 500-meter radius around a subway station is considered a reasonable transfer distance. If the time interval between adjacent chain nodes in the refined structure exceeds the preset transfer time threshold or the spatial distance between adjacent chain nodes exceeds the preset walking distance threshold, it indicates a conflict between the candidate and the actual transfer conditions in the refined structure, and one conflict count is added.

[0044] If the candidate connection structure is of the "missing boarding / alighting" type, then check whether the positional span between the supplemented boarding / alighting station in the refined structure and the subsequent card-swiping event node exceeds a preset connection span threshold. The connection span threshold is set based on the bus route stop sequence interval and the urban road network, typically limited to 3 to 5 stops or 2 kilometers, for example, set to 5 stops or a straight-line distance of 2 kilometers. If the positional span between the supplemented boarding / alighting station in the refined structure and the subsequent card-swiping event node exceeds the preset connection span threshold, it indicates that the supplemented station cannot be reasonably connected to the subsequent node, and one conflict count is added.

[0045] If the candidate connection structure is a short-stay type, then check whether the time interval between adjacent chain nodes in the refined structure exceeds the preset short-stay time threshold, or whether the spatial locations of adjacent chain nodes belong to different stations. The short-stay time threshold is usually set to 10 to 15 minutes, indicating that the user stays at the same station or within a very small area before continuing their journey. If the time interval between adjacent chain nodes in the refined structure exceeds the preset short-stay time threshold or the spatial locations of adjacent chain nodes belong to different stations (different traffic location codes), it indicates that the actual process is not a short-stay, and a conflict count is added.

[0046] If the candidate transition structure is an independent travel start type, then the pre- and post-travel nodes in the refined structure are checked to see if they conform to the preset traffic continuity relationship. The preset traffic continuity relationship is determined by at least one of the following parameter dimensions: In the time dimension, the time interval is within a reasonable transfer time window, with the lower limit being the minimum transfer time threshold (e.g., 2 minutes, used to exclude false judgments caused by consecutive card swipes on the same trip) and the upper limit being the maximum transfer time threshold (e.g., 60 minutes, used to exclude independent travel caused by long stays); In the spatial dimension, the spatial distance is less than or equal to the preset maximum walking distance threshold (e.g., 800 meters to 1500 meters, set according to the walking accessibility between stations, 800 meters can be used in the urban core area, and 1500 meters can be used in the suburbs); In the path accessibility dimension, there is at least one bus or subway line that can connect the two locations, and there is a trip that can complete the path within the time window. If the pre-set traffic continuity relationship exists between the pre- and post-travel nodes in the refined structure, it means that the candidate type of independent travel start conflicts with the facts, and then one conflict count is added.

[0047] If the candidate receiving structure is an abnormal card swipe type, then check whether the spatiotemporal correlation characteristics between the chain nodes and the card swipe event nodes in the refined structure meet the preset correlation conditions. The preset correlation conditions include: the reasonable time difference between the card swipe time of the node and the adjacent node should be within the normal operating time; the node locations should belong to the same transportation network and there should be no logical contradictions (for example, the previous node is on a bus, and the next node appears at a subway station 50 kilometers away 10 minutes later, and there is no other mode of transportation that can reach it). If the spatiotemporal correlation characteristics between the chain nodes and the card swipe event nodes in the refined structure meet the preset correlation conditions, it means that the card swipe event is not actually abnormal, and then one conflict count is added.

[0048] Through the above item-by-item comparison, the number of structural conflicts for each candidate receiving structure is statistically analyzed and directly used as the degree of structural conflict. A negative correlation mapping is then performed based on the degree of structural conflict to obtain the acceptance credibility of each candidate receiving structure. The meaning of the negative correlation mapping is: the more conflicts, the lower the acceptance credibility; the fewer conflicts, the higher the acceptance credibility. A commonly used mapping method is to first add 1 to the number of conflicts, and then take the reciprocal transformation to obtain the acceptance credibility. Adding 1 to the denominator avoids the denominator being zero, and when the number of conflicts is zero, the acceptance credibility is 1; the larger the number of conflicts, the closer the credibility approaches 0. Other feasible negative correlation mappings include exponential decay or linear decay, etc. This scheme prefers the reciprocal transformation because it is simple to calculate and requires no additional parameters.

[0049] S4. Select the candidate acceptance structure with the highest acceptance credibility as the optimal acceptance structure between adjacent card swipe event nodes to generate the user travel chain.

[0050] The aforementioned steps quantify the consistency between each candidate structure and the refined structure. To construct a complete and reasonable user travel chain, the most reliable structure needs to be selected from multiple candidate structures as the actual connection relationship between adjacent card-swiping event nodes. Based on this, card-swiping event nodes are merged, split, or supplemented. The optimal connection structure refers to the candidate connection structure with the highest connection reliability, representing the best inference of the actual traffic process between adjacent nodes. The user travel chain is an ordered sequence consisting of a series of card-swiping event nodes arranged in chronological order and supplementary chain nodes (such as inferred drop-off stations). It is used to represent the user's complete travel trajectory within the statistical period, including the starting point, ending point, transfer point, and splitting points between different trips for each trip.

[0051] In some implementations, if multiple candidate structures have the same highest credibility, they are selected according to a preset priority rule. For example, the priority rule prioritizes the "transfer" type, followed by the "missing boarding / alighting" type, then the "short stop" type, and finally the "independent trip start" type. The "abnormal card swipe" type is only selected if its credibility is significantly higher than other types; otherwise, other types are prioritized to avoid excessive data removal. After determining the optimal connecting structure, the corresponding trip chain generation operation is performed according to its type. If the optimal connection structure is a transfer type, it means that two adjacent card swipe event nodes belong to the same continuous trip. They are connected by a transfer in the middle, which merges the adjacent card swipe event nodes into the same trip chain and records the transfer point. The location of the transfer point between the two nodes is recorded (usually the location of the next card swipe event node or the station code corresponding to the transfer channel).

[0052] If the optimal connection structure is the type of missing boarding / alighting, it means that there is a missing alighting record between the previous card swiping event node (e.g., boarding a bus) and the next card swiping event node (e.g., entering a subway station or boarding another bus). The missing boarding / alighting station will be inserted as a new node into the travel chain. The insertion position is located between the two card swiping event nodes. The new node carries the tag "supplemented alighting point" to distinguish it from the real card swiping node.

[0053] If the optimal connection structure is a short stay type, it means that the user experienced a short stay between two card swipe event nodes (such as transferring or waiting at the same station), but the two card swipe behaviors are still part of the same trip. The adjacent card swipe event nodes are merged into the same trip chain without inserting any new nodes, and only the node order is kept continuous.

[0054] If the optimal connection structure is an independent trip start type, it means that the next card swipe event node belongs to a completely new trip and has no connection relationship with the previous card swipe event node. The next card swipe event node is used as the starting point of a new trip chain to avoid erroneously merging two independent trips into one trip.

[0055] If the optimal connection structure is an abnormal card swipe type, it indicates that the card swipe event node has problems such as duplicate charges, time offset, or abnormal location, and should not be included in the normal travel chain. The corresponding card swipe event node is then removed. After removal, the intervals before and after the node are merged into a new interval to be explained. The connection analysis is then performed again on this new interval. At the same time, it is determined whether the number of consecutive re-analysis of the current interval has reached the preset limit. The limit is set based on the statistical distribution law that consecutive abnormal events in actual traffic card swipe data are low-probability isolated events, and the principle of balancing computational efficiency that the marginal improvement of the travel chain restoration degree after exceeding this number approaches zero. It is preferably set to 3 times. In actual traffic card swipe data, if an interval still cannot form a reasonable connection relationship after 3 consecutive abnormal removals, it usually indicates that the data segment has irreparable quality problems (such as multiple consecutive re-transmission records or accumulated device time offset). Continuing to re-analyze will lead to ineffective iteration. Therefore, when a reasonable result cannot be obtained after the 3rd re-analysis, the re-analysis is terminated, and the next card swipe node is used as the starting point of the new travel chain to restore the sequential processing process. If the limit is reached, stop re-analysis and use the next card swipe node as the starting point of a new travel chain to resume the sequential processing.

[0056] For abnormal card-swiping event nodes that cannot be directly removed due to data quality reasons but whose statistical contribution needs to be reduced (including nodes whose location information is supplemented by vehicle positioning and route station sequence, nodes whose time offset exceeds the preset offset threshold, and other records other than the earliest record retained in the duplicate deduction records generated by the same device within the preset duplicate detection time window), a weight label is added after the travel chain is generated. This weight is used for counting weighting in subsequent passenger flow origin-destination statistics and transfer relationship statistics. Specifically, the weight coefficient is set according to the abnormality type: for the earliest record retained for duplicate deduction, the weight coefficient is set to 0.5; for nodes whose time offset exceeds the preset offset threshold (e.g., 5 minutes), the weight coefficient is set to 0.3; for nodes whose location information is supplemented by vehicle positioning and route station sequence, the weight coefficient is set to 0.8. If the same node has multiple abnormal labels at the same time, the minimum weight coefficient corresponding to each abnormality type is taken as the final weight.

[0057] Through the above processing, severely abnormal nodes are completely removed and repairable abnormal nodes are retained with reduced weight. This ensures the integrity of the travel chain structure while avoiding significant deviations in statistical results caused by abnormal data.

[0058] Through the above operations, all adjacent node pairs in the card swipe event node sequence are processed sequentially, ultimately generating one or more complete travel chains for each user. The node sequence within each travel chain is continuous in time, reachable in space, and logically self-consistent, containing both actual card swipe event nodes and limited, refined chain nodes.

[0059] Furthermore, after generating user travel chains, the travel chains generated by the same user within the same operating cycle (defined according to analysis needs, such as a calendar day or a statistical week) are arranged chronologically to form a user travel chain sequence. Based on the user travel chain sequence, the starting point (the position of the first node), ending point (the position of the last node), transfer point (the position marked as a transfer point in the travel chain), mode of transport combination (the sequence of modes of transport at each node in the travel chain), and time window (starting time and ending time) of each travel chain are extracted for at least one of the following travel big data analyses: Origin-destination passenger flow statistics: Statistically count the total number of trips from the same origin area to the same destination area within a specified time window, and generate an OD (origin-destination) matrix to analyze passenger flow distribution patterns and traffic demand intensity.

[0060] Transfer relationship statistics: Statistics on the number and volume of transfers between different routes and different modes of transportation, identifying key transfer hubs and transfer pressure nodes, providing a basis for route optimization and capacity allocation.

[0061] Commuting pattern identification: Identify fixed origin-destination pairs and fixed travel time windows that recur for the same user in multiple operating cycles, and mark them as commuting trips for analysis of residents' commuting characteristics and work-residence distribution.

[0062] Regional passenger flow change analysis: Divide regions according to administrative divisions, traffic zones or station influence ranges, and count the passenger flow entering the station, exiting the station and the passenger flow between regions in each region during a specified time period to identify peak passenger flow periods and hot spots.

[0063] Based on the above technical solution, the traffic process between adjacent card-swiping event nodes is abstracted into candidate receiving structures, and further mapped into chain structures for difference analysis to obtain the chain structure refinement requirement. This allows the supplementation of traffic process information to be implemented in a targeted manner based on the degree of difference at the chain structure level. On this basis, by comparing the degree of structural conflict between each candidate receiving structure and the refined structure obtained after supplementation, the candidate receiving structure with the highest receiving credibility is selected as the optimal receiving structure to generate the user travel chain. This improves the construction of the travel chain from a simple time sequence splicing to a refined processing based on chain structure difference analysis and structural conflict comparison, so that the travel chain generation process can reflect the divergence impact of different candidate receiving structures at the chain structure level.

Claims

1. A method for analyzing travel big data based on transportation card swipe information, characterized in that, include: Generate a sequence of card swipe event nodes based on traffic card swipe records; List multiple candidate connection structures between adjacent card swipe event nodes, and determine the degree of refinement requirement for the chain structure between adjacent card swipe event nodes based on the differences between the chain structures corresponding to the multiple candidate connection structures. The traffic process information between adjacent card-swiping event nodes is supplemented based on the refined demand degree of the chain structure to obtain a refined structure. The acceptance credibility of each candidate acceptance structure is determined based on the degree of structural conflict between each candidate acceptance structure and the refined structure. The candidate acceptance structure with the highest acceptance credibility is selected as the optimal acceptance structure between adjacent card swipe event nodes to generate the user travel chain.

2. The travel big data analysis method based on transportation card swiping information according to claim 1, characterized in that, List several candidate succession structures between adjacent card swipe event nodes, including: Based on the time interval, spatial location relationship, route connection relationship and mode of transportation conversion relationship between adjacent card swipe event nodes, candidate succession structures are generated, including at least one type of transfer, missing boarding / alighting, short stop, start of independent travel and abnormal card swipe.

3. The travel big data analysis method based on transportation card swiping information according to claim 2, characterized in that, Based on the differences between the chain structures corresponding to multiple candidate receiving structures, the required level of refinement of the chain structure between adjacent card swipe event nodes is determined, including: Extract at least one chain node position from each candidate receiving structure, including boarding / alighting stations, transfer points, and trip chain disconnection points, and map it to the corresponding chain structure. Count the number of duplicate chains corresponding to multiple candidate acceptance structures to determine the node acceptance divergence degree between adjacent card swipe event nodes. Statistically analyze the differences in chain node positions among the chain structures corresponding to multiple candidate successor structures; The degree of refinement required for the chain structure is determined by combining the degree of divergence of the nodes and the differences in the positions of the chain nodes.

4. The travel big data analysis method based on transportation card swiping information according to claim 2, characterized in that, Based on the refined demand degree of the chain structure, the traffic process information between adjacent card-swiping event nodes is supplemented to obtain a refined structure, including: If the required refinement of the chain structure is greater than the preset threshold, then the station topology data, line station sequence data, transfer channel data or train information are called to supplement the boarding and alighting stations, transfer points and disconnection points of the travel chain between adjacent card swipe event nodes to form the refined structure. If the required refinement degree of the chain structure is less than or equal to the preset requirement threshold, then any candidate receiving structure will be used as the refined structure.

5. The travel big data analysis method based on transportation card swiping information according to claim 4, characterized in that, The reliability of each candidate receiving structure is determined based on the degree of structural conflict between each candidate receiving structure and the refined structure, including: The chain node positions of the refined structure are compared item by item according to the type of each candidate receiving structure, the number of conflicts of the chain node positions is counted, and the degree of conflict of the structure is determined. Based on the degree of structural conflict, a negative correlation mapping is performed to obtain the acceptance credibility of each candidate acceptance structure.

6. The travel big data analysis method based on transportation card swiping information according to claim 5, characterized in that, The number of collisions at the chain node positions is counted, including: If the candidate receiving structure is a transfer type, and the time interval between adjacent chain nodes in the refined structure is greater than a preset transfer time threshold or the spatial distance between adjacent chain nodes is greater than a preset walking distance threshold, then the number of conflicts is accumulated. If the candidate connection structure is of the type with missing boarding / alighting points, and the position span between the boarding / alighting station supplemented in the refined structure and the subsequent card swipe event node is greater than the preset connection span threshold, then the number of conflict counts is accumulated. If the candidate receiving structure is a short stay type, and the time interval between adjacent chain nodes in the refined structure is greater than the preset short stay time threshold or the spatial location of adjacent chain nodes belongs to different stations, then the number of conflicts is accumulated. If the candidate receiving structure is an independent trip start type, and there is a preset traffic continuation relationship between the preceding and following chain nodes in the refined structure, then the number of conflict counts is accumulated. If the candidate receiving structure is an abnormal card swipe type, and the spatiotemporal correlation characteristics between the chain node and the card swipe event node in the refined structure meet the preset correlation conditions, then the number of conflicts is accumulated.

7. The travel big data analysis method based on transportation card swiping information according to claim 6, characterized in that, The candidate connection structure with the highest connection credibility is selected as the optimal connection structure between adjacent card swipe event nodes to generate a user travel chain, including: If the optimal connection structure is a transfer type, the adjacent card swipe event nodes are merged into the same travel chain and the transfer point is recorded; If the optimal connection structure is a type of missing boarding / alighting points, the missing boarding / alighting points will be inserted as new nodes into the travel chain. If the optimal connection structure is a short-stay type, the adjacent card swipe event nodes will be merged into the same travel chain; If the optimal connection structure is an independent trip start type, the next card swipe event node will be used as the starting point of the new trip chain; If the optimal receiving structure is an abnormal card swipe type, the corresponding card swipe event node will be removed.

8. The travel big data analysis method based on traffic card swiping information according to claim 1, characterized in that, Generate a sequence of card swipe event nodes based on the card swipe records, including: Obtain the original card swipe records, process each original card swipe record, remove records that are missing user identifiers or card swipe times, and fill in records that are missing location information based on vehicle location and route station sequence; Each processed original card swipe record is converted into a card swipe event node, grouped according to user identifier, and arranged by swipe time within the group to obtain the card swipe event node sequence.

9. The travel big data analysis method based on traffic card swiping information according to claim 1, characterized in that, After generating the user's travel chain, it also includes: Arrange the travel chains generated by the same user within the same operating cycle in chronological order to form a user travel chain sequence; Based on the user travel chain sequence, the starting point, ending point, transfer point, mode of transport combination and time window of each travel chain are extracted to perform passenger flow origin and destination flow statistics, transfer relationship statistics, commuting pattern identification or regional passenger flow change analysis.

10. A travel big data analysis system based on traffic card swiping information, characterized in that, include: The node sequence generation module is used to generate a sequence of card swipe event nodes based on traffic card swipe records; The detailed requirement determination module is used to list multiple candidate receiving structures between adjacent card swiping event nodes, and determine the degree of detailed requirement of the chain structure between adjacent card swiping event nodes based on the differences between the chain structures corresponding to multiple candidate receiving structures. The credibility analysis module is used to supplement the traffic process information between adjacent card swiping event nodes based on the refined demand of the chain structure to obtain a refined structure, and to determine the acceptance credibility of each candidate acceptance structure based on the degree of structural conflict between each candidate acceptance structure and the refined structure. The travel chain generation module is used to select the candidate connection structure with the highest connection credibility as the optimal connection structure between adjacent card swipe event nodes, and generate a user travel chain.

Images