Behavior portrait generation method, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention provides a behavior portrait generation method, electronic equipment, a storage medium and a program product. The method comprises the following steps: performing spatio-temporal clustering processing based on a plurality of historical track data of a user to obtain key interest points; extracting a plurality of path fragments from the plurality of pieces of historical track data; determining preference path data according to the frequency characteristics of the plurality of path segments; determining abnormal path data according to the multi-dimensional dynamic behavior data of the plurality of path segments; and determining a behavior portrait of the user according to the key interest point, the preference path data and the abnormal path data. The method is used for improving the accuracy of the behavior portrait.

Description

Technical Field

[0001] This application relates to the field of user profile generation, and in particular to a behavioral profile generation method, electronic device, storage medium, and program product. Background Technology

[0002] Intelligent navigation services provide personalized navigation recommendations based on user behavior profiles; therefore, the accuracy of these behavioral profiles has a significant impact on intelligent navigation services.

[0003] In related technologies, users' static attributes are usually mined to build user behavior profiles. For example, user behavior profiles are built based on basic user information and user-defined travel patterns. However, this can lead to discrepancies between the behavior profiles and the user's actual behavioral characteristics, resulting in low accuracy of the behavior profiles. Summary of the Invention

[0004] This application provides a behavioral profile generation method, electronic device, storage medium, and program product to improve the accuracy of behavioral profiles.

[0005] In a first aspect, embodiments of this application provide a method for generating behavioral profiles, including:

[0006] Spatiotemporal clustering is performed on multiple historical trajectory data of users to obtain key points of interest; multiple path segments are extracted from the multiple historical trajectory data; preferred path data is determined based on the frequency characteristics of multiple path segments; abnormal path data is determined based on the multidimensional dynamic behavior data of multiple path segments; and user behavior profile is determined based on key points of interest, preferred path data, and abnormal path data.

[0007] Secondly, embodiments of this application provide a behavior profile generation apparatus, comprising:

[0008] The spatiotemporal clustering module is used to perform spatiotemporal clustering based on multiple historical trajectory data of users to obtain key points of interest;

[0009] The extraction module is used to extract multiple path segments from multiple historical trajectory data.

[0010] The preference determination module is used to determine the preferred path data based on the frequency characteristics of multiple path segments;

[0011] The anomaly detection module is used to determine abnormal path data based on the multidimensional dynamic behavior data of multiple path segments.

[0012] The behavior profile determination module is used to determine the user's behavior profile based on key points of interest, preference path data, and abnormal path data.

[0013] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory, causing the processor to perform:

[0014] Spatiotemporal clustering is performed on multiple historical trajectory data of users to obtain key points of interest; multiple path segments are extracted from the multiple historical trajectory data; preferred path data is determined based on the frequency characteristics of multiple path segments; abnormal path data is determined based on the multidimensional dynamic behavior data of multiple path segments; and user behavior profile is determined based on key points of interest, preferred path data, and abnormal path data.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement:

[0016] Spatiotemporal clustering is performed on multiple historical trajectory data of users to obtain key points of interest; multiple path segments are extracted from the multiple historical trajectory data; preferred path data is determined based on the frequency characteristics of multiple path segments; abnormal path data is determined based on the multidimensional dynamic behavior data of multiple path segments; and user behavior profile is determined based on key points of interest, preferred path data, and abnormal path data.

[0017] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements:

[0018] Spatiotemporal clustering is performed on multiple historical trajectory data of users to obtain key points of interest; multiple path segments are extracted from the multiple historical trajectory data; preferred path data is determined based on the frequency characteristics of multiple path segments; abnormal path data is determined based on the multidimensional dynamic behavior data of multiple path segments; and user behavior profile is determined based on key points of interest, preferred path data, and abnormal path data.

[0019] The behavioral profile generation method, electronic device, storage medium, and program product provided in this application's embodiments determine key interest points through spatiotemporal clustering, making these key interest points densely distributed trajectory points in both spatial and temporal dimensions, thus improving the accuracy of key interest point determination. By analyzing the frequency characteristics of path segments, preference path data is determined, distinguishing between occasional and long-term preference paths, thereby improving the accuracy of preference path data. Furthermore, by analyzing the multidimensional dynamic behavioral data of path segments, abnormal path data is determined, making the detection of abnormal behavior more comprehensive and effectively detecting abnormal path data across multiple dimensions. This method dynamically mines key interest points, preference path data, and abnormal path data from multiple historical trajectory data of the user, improving the real-time nature of the behavioral profile, ensuring that the behavioral profile matches the user's actual behavioral characteristics, and enhancing the effectiveness and accuracy of the behavioral profile. Attached Figure Description

[0020] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0021] Figure 1 A schematic diagram of a scenario for the behavior profile generation method provided in this application;

[0022] Figure 2 A schematic diagram of the thermal bubble diagram for spatiotemporal clustering provided in this application;

[0023] Figure 3 A schematic diagram of the behavioral profile generation system provided in this application;

[0024] Figure 4 A schematic diagram of the behavioral profile generation device provided in this application;

[0025] Figure 5 A schematic diagram of the structure of the electronic device provided in this application.

[0026] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0027] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0028] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0029] Figure 1 Flowchart of the behavioral profile generation method provided in this application Figure 1 Behavioral profiling generation methods can be applied to electronic devices, which can be terminals or servers; for example... Figure 1 As shown, the behavioral profile generation method includes:

[0030] S101. Spatiotemporal clustering is performed based on multiple historical trajectory data of users to obtain key points of interest.

[0031] Historical trajectory data includes at least a link sequence and a trajectory point sequence. The trajectory point sequence includes multiple trajectory points, which are represented by their location and timestamp. Historical trajectory data may also include user identifier, vehicle identification number (VIN), and trip identifier. The location of the trajectory points can be represented by their longitude and latitude.

[0032] Key points of interest can be users' frequently used starting points or destinations; in practical applications, key points of interest can be home, company, school, etc.

[0033] Optionally, historical trajectory data of floating cars from anonymous users can be collected. This historical trajectory data includes user identifier, vehicle identification number, trip identifier, trajectory point longitude, trajectory point latitude, and trajectory point timestamp, as exemplarily shown in Table 1.

[0034] Table 1

[0035]

[0036] The historical trajectory data of floating cars is collected frequently and covers a wide range, but there may be problems such as sensor errors and positioning drift. In order to improve the quality of trajectory data, the historical trajectory data of floating cars is cleaned and preprocessed. Through cleaning and preprocessing, abnormal trajectories are removed, duplicate trajectory points and flying points are filtered out, trajectory segmentation, tunnel stitching, stop point identification, map matching, etc., to obtain the user's historical trajectory.

[0037] For example, the user's historical trajectory data is shown in Table 2.

[0038] Table 2

[0039]

[0040] Specifically, multiple historical trajectory data of the user are obtained, and the starting point and the ending point are extracted from the multiple historical trajectory data to obtain multiple starting points and multiple ending points of the user; based on the trajectory point locations and trajectory point timestamps of multiple starting points and multiple ending points, spatiotemporal clustering is performed to obtain key points of interest.

[0041] It should be noted that spatiotemporal clustering processing makes key points of interest densely distributed trajectory points in both spatial and temporal dimensions, thus improving the accuracy of identifying key points of interest.

[0042] S102. Extract multiple path segments from multiple historical trajectory data.

[0043] The path segment includes one or more links.

[0044] Historical trajectory data includes road segment sequences. Optionally, for each historical trajectory data, each road segment in the road segment sequence included in the historical trajectory data is taken as a path segment.

[0045] Optionally, multiple path segments can be extracted from multiple historical trajectory data, including: for each historical trajectory data, selecting path segments from the road segment sequence included in the historical trajectory data through a sliding window.

[0046] Specifically, the length of the sliding window, i.e. the number of road segments included in the path segment, the length of the sliding window, and the sliding length of the sliding window can all be determined according to actual needs.

[0047] For example, user The historical trajectory data includes the sequence of road segments represented as follows: ,in, It consists of multiple links sorted by time; a sliding window method is used to extract links of length [missing information]. The path segment, represented as: .

[0048] Understandably, this involves multiple path segments, including those extracted from each historical trajectory data through a sliding window.

[0049] In the above embodiments, by sliding a window to select path segments in historical trajectory data, user preference analysis and behavior analysis can be performed based on some segments of historical trajectory data, which can achieve more granular behavior profiles and improve the accuracy of behavior profiles.

[0050] S103. Determine the preferred path data based on the frequency characteristics of multiple path segments.

[0051] It should be noted that multiple path segments are determined based on multiple historical trajectory data. For two historical trajectory data with a high degree of overlap, path segments with the same location information can be extracted based on these two historical trajectory data. Therefore, there are path segments with the same location information among multiple path segments.

[0052] The frequency feature of the path segment is used to reflect how frequently path segments with the same location information appear in multiple historical trajectory data. For example, the frequency feature includes frequency count and frequency frequency, which are used to reflect the number of times and the frequency of occurrence of path segments with the same location information in multiple historical trajectory data.

[0053] Preference path data is the relevant data of preference path segments; the relevant data may include the time and frequency characteristics of preference path segments; preference path segments are path segments that users frequently take, and can also be called familiar path segments.

[0054] Optionally, based on multiple path segments, the frequency characteristics of each path segment are determined, and the path segments whose flatness characteristics meet the preset preference conditions are taken as preferred path segments. Based on the relevant data of the preferred path segments, preferred path data is determined.

[0055] Optionally, multiple path segments are divided according to time to obtain path segments belonging to multiple preset time periods; based on the multiple path segments in each preset time period, the frequency characteristics of each path segment in the preset time period are determined, and the path segments whose frequency characteristics meet preset preference conditions are taken as the preferred path segments in the preset time period; based on the relevant data of the preferred path segments in multiple preset time periods, the preferred path data is determined.

[0056] S104. Based on the multidimensional dynamic behavior data of multiple path segments, determine the abnormal path data.

[0057] The multidimensional dynamic behavior data includes: the associated behavior data generated by the vehicle driving on the path segment, and the trajectory data before the path segment.

[0058] Abnormal path data refers to the relevant data of abnormal path segments; this relevant data may include multi-dimensional dynamic behavior data and abnormal types of abnormal path segments.

[0059] Specifically, for each path segment, based on the time and location information of the path segment, the associated behavioral data of the path segment is obtained from the user's historical driving data and vehicle-road data; and the previous path trajectory data before the path segment is obtained from the historical trajectory data containing the path segment.

[0060] Based on the associated behavior data, determine whether the path segment violates traffic rules and / or safe driving rules. If so, the path segment is regarded as an abnormal path segment. Based on the data of traffic rule and / or safe driving rule violations in the associated behavior data, determine the preferred path data of the abnormal path segment.

[0061] By using a neural network model, the path segment and the preceding trajectory data are analyzed to determine the abnormal score of the path segment when the preceding trajectory data exists. If the abnormal score determines that the path segment is an abnormal path segment, the abnormal path data of the abnormal path segment is determined based on the preceding trajectory data.

[0062] S105. Based on key points of interest, preference path data, and abnormal path data, determine the user's behavioral profile.

[0063] Optionally, the behavioral profile includes key points of interest, preference path data, and abnormal path data.

[0064] Optionally, a basic static behavior profile of the user is determined, a static behavior profile is determined based on the basic static behavior profile and key points of interest, and a dynamic behavior profile is determined based on preference path data and abnormal path data. The user's behavior profile includes a static behavior profile and a dynamic behavior profile.

[0065] Determining a user's basic static behavior profile can be done by using the user's trajectory data and driving data. For example, the basic static behavior profile includes: average number of trips per day, differences between weekday and weekend trips, high-frequency departure / arrival time periods, daily activity range, commuting distance, commuting time, average driving speed, and average driving mileage.

[0066] Optionally, user behavior profiles can be written into a behavior profile database. User behavior profiles can be stored in time partitions for archiving, rollback, and viewing of historical versions, and also for incremental update management.

[0067] Optionally, user behavior profiles can be determined periodically based on historical trajectory data within a historical period. For example, every week, the user's historical trajectory data from the past three months can be re-acquired, and the user's behavior profile can be determined based on the re-acquired historical trajectory data. The current behavior profile can be used to overwrite the behavior profiles stored in the profile database, or the behavior profiles stored in the profile database can be updated, forming a complete closed-loop update process to ensure that the behavior profile can reflect the user's latest behavior. When merging and updating the currently determined behavior profile and the historical behavior profile, the matching historical behavior profile can be determined based on the tags of the currently determined behavior profile (such as user ID), and then the merging and updating can be performed, which improves the accuracy and efficiency of behavior profile updates.

[0068] The behavioral profile generation method provided in this application determines key interest points (PIs) through spatiotemporal clustering, making them densely distributed trajectory points in both spatial and temporal dimensions, thus improving the accuracy of PI determination. It also determines preference path data through the frequency characteristics of path segments, distinguishing between occasional and long-term preference paths, thereby improving the accuracy of preference path data. Furthermore, it identifies abnormal path data through multi-dimensional dynamic behavioral data of path segments, making the detection of abnormal behavior more comprehensive and effectively detecting abnormal path data across multiple dimensions. This method dynamically mines users' key interest points, preference path data, and abnormal path data from multiple historical trajectory data, improving the real-time nature of behavioral profiles and ensuring that the behavioral profiles match the users' actual behavioral characteristics, thereby enhancing the effectiveness and accuracy of behavioral profiles.

[0069] In some embodiments, spatiotemporal clustering is performed on multiple historical trajectory data of a user to obtain key points of interest, including: extracting the start and end points from the multiple historical trajectory data of the user to obtain multiple candidate points of interest; determining the spatiotemporal weights of the multiple candidate points of interest based on their time and location; performing spatiotemporal clustering on the multiple candidate points of interest to obtain multiple clusters; selecting a target cluster from the multiple clusters based on the spatiotemporal weights of the candidate points of interest included in the multiple clusters; and determining the key points of interest based on the location of the candidate points of interest included in the target cluster.

[0070] Among them, the spatiotemporal weight can reflect the similarity between the time of the candidate point of interest and the preset time, and between the location of the candidate point of interest and the location of the neighboring points of interest.

[0071] Specifically, for each historical trajectory data, the starting point and the ending point are extracted from the sequence of trajectory points included in the historical trajectory data. In this way, multiple starting points and multiple ending points can be extracted from multiple historical trajectory data, and multiple starting points and multiple ending points are all used as candidate points of interest.

[0072] For each candidate point of interest, the spatiotemporal weight of the candidate point of interest is determined based on its time (time stamp of the trajectory point), location (longitude and latitude of the trajectory point), preset time, and the location of neighboring points of interest.

[0073] Optionally, spatiotemporal clustering is performed on multiple candidate interest points to obtain multiple clusters. Each cluster includes multiple candidate interest points with dense temporal and spatial distribution. For each cluster, the spatiotemporal score of the cluster is determined based on the spatiotemporal weights of the multiple candidate interest points in the cluster. The cluster with the largest spatiotemporal score among the multiple clusters is taken as the target cluster. The location of the key interest point is obtained by weighted summation based on the location of the candidate interest points included in the target cluster.

[0074] Optionally, multiple candidate points of interest extracted from multiple historical trajectory data, including multiple candidate starting points and multiple candidate ending points, can be subjected to spatiotemporal clustering processing for multiple candidate starting points, and the obtained key points of interest can be used as key starting points; and spatiotemporal clustering processing can be performed for multiple candidate ending points, and the obtained key points of interest can be used as key ending points.

[0075] Optionally, after extracting multiple candidate points of interest (POIs), they are divided into multiple first candidate POIs, multiple second candidate POIs, multiple third candidate POIs, and multiple fourth candidate POIs based on their time and location. For example, the first candidate POIs might be candidate starting points for the morning rush hour, the second candidate POIs might be candidate ending points for the morning rush hour, the third candidate POIs might be candidate starting points for the evening rush hour, and the fourth candidate POIs might be candidate ending points for the evening rush hour. Spatiotemporal clustering is then performed on the first candidate POIs, and the resulting key POIs are used as key starting points for the morning rush hour; spatiotemporal clustering is performed on the second candidate POIs, and the resulting key POIs are used as key ending points for the morning rush hour; spatiotemporal clustering is performed on the third candidate POIs, and the resulting key POIs are used as key starting points for the evening rush hour; spatiotemporal clustering is performed on the fourth candidate POIs, and the resulting key POIs are used as key ending points for the evening rush hour. In practical applications, the key starting points, key ending points, key starting points, and key ending points for the morning and evening rush hours can be used to determine a user's "home," "company," or other points of interest.

[0076] In practical applications, the spatiotemporal DBSCAN algorithm can be used for spatiotemporal clustering. The spatiotemporal DBSCAN algorithm can also be referred to as ST-DBSCAN (Spatial Temporal Density-Based Spatial Clustering of Applications with Noise), which is a density-based spatiotemporal noise application spatial clustering algorithm.

[0077] In the above embodiments, the spatiotemporal weights of multiple candidate interest points are determined based on their time and location. The multiple candidate interest points are then processed by spatiotemporal clustering to obtain multiple clusters. Based on the spatiotemporal weights of the candidate interest points included in the multiple clusters, the target cluster is determined, thereby identifying the key interest points, which improves the robustness and accuracy of key interest point identification.

[0078] In some embodiments, determining the spatiotemporal weights of multiple candidate interest points based on their time and location includes: for each candidate interest point, determining neighboring interest points among the multiple candidate interest points; determining a time weight based on a preset time and the time of the candidate interest points; determining a spatial weight based on the location of the domain interest points and the location of the candidate interest points; and determining the spatiotemporal weights of the candidate interest points based on the time weights and the spatial weights.

[0079] The preset time can be set according to actual needs. In practical applications, the preset time can be either the morning peak time or the evening peak time.

[0080] Specifically, for each candidate point of interest, a preset time is determined based on the time of the candidate point of interest. For example, if the time of the candidate point of interest falls within the morning peak period, then the morning peak period time is used as the preset time; if the event of the candidate point of interest falls within the evening peak period, then the evening peak period time is used as the preset time. That is, the time of the candidate point of interest is determined between the morning peak period and the evening peak period. More closely, as a preset time Based on candidate points of interest Time and preset time The time weights are determined using Gaussian decay. By attenuating the time weights of candidate interest points using Gaussian decay, the time weights of candidate interest points that are close to the preset time are increased, which can effectively filter out candidate interest points that do not stay at high frequencies.

[0081] Based on the locations of multiple candidate interest points, neighborhood interest points are determined from among the candidate interest points, denoted as […]. Based on the location of candidate points of interest Location of neighboring points of interest Spatial weights are determined by kernel density estimation; and spatial clustering of real dwell points is enhanced by combining the neighborhood distribution of candidate interest points.

[0082] The temporal and spatial weights of the candidate interest points are weighted and summed to obtain the spatiotemporal weights of the candidate interest points. In this embodiment, the spatiotemporal weights of multiple candidate interest points are determined based on their events and locations, so that key interest points can be determined subsequently based on the spatiotemporal weights. This can effectively reduce the impact of infrequent dwell points and occasional abnormal dwell points on the clustering results, making the determined key interest points consistent with the user's actual behavior and improving the accuracy of key interest points.

[0083] Optionally, the location of the key interest point is obtained by weighted summation based on the locations of the candidate interest points included in the target cluster, including: weighted summation of the locations of the candidate interest points included in the target cluster based on their spatial weights.

[0084] For example, the time weight is calculated by formula (1); the spatial weight is calculated by formula (2); and the spatiotemporal weight is calculated by formula (3).

[0085] Formula (1): ;

[0086] Formula (2): ;

[0087] Formula (3): ;

[0088] in, It is time weighting. Candidate points of interest Time, It is a preset time. It is a preset time tolerance, used to control time sensitivity; It is spatial weight. Indicates the location of the candidate point of interest. Indicate candidate points of interest Neighborhood points of interest It is the location of the neighboring interest point. These are preset spatial scale parameters used to control spatial sensitivity. Candidate points of interest and neighboring points of interest The Euclidean distance between them; It is the first preset parameter. It is the second preset parameter. , Candidate points of interest The spatiotemporal weights.

[0089] The time tolerance, space tolerance, first preset parameter and second preset parameter can all be set according to actual needs. In this application embodiment, the specific values ​​of the time tolerance, space tolerance, first preset parameter and second preset parameter are not limited.

[0090] For each cluster, the spatiotemporal score of the cluster is obtained by averaging the spatiotemporal weights of the multiple candidate interest points included in the cluster, as shown in formula (4).

[0091] Formula (4): ;

[0092] in, It is a cluster The spatiotemporal fraction, It is the number of candidate interest points in the cluster. It is the spatiotemporal weight of candidate interest points in the cluster.

[0093] After obtaining the spatiotemporal score of each cluster, the cluster with the highest spatiotemporal score is selected as the target cluster. .

[0094] Based on the spatial weights of the candidate interest points included in the target cluster, the positions of the candidate interest points included in the target cluster are weighted and summed to obtain the positions of the key interest points, as shown in formula (5).

[0095] Formula (5): ;

[0096] in, It is the target cluster. These are candidate points of interest within the target cluster. It is the longitude of the candidate point of interest. It is the dimension of the candidate points of interest. Candidate points of interest Spatial weights, This represents a weighted summation of the latitude and longitude coordinates of all candidate points of interest in the target cluster. It is the sum of the spatial weights of all candidate interest points in the target cluster, used to achieve normalization. It is the center of the target cluster and also the location of the key point of interest.

[0097] For example, such as Figure 2 As shown, spatiotemporal clustering is performed on multiple candidate points of interest for users. A spatiotemporal access heatmap is determined based on the clustering results, where the horizontal axis represents hours (time of day), the vertical axis represents the cluster number, and the bubble size represents the number of candidate points of interest in each cluster. Figure 2It can be determined that, based on clustering the user's multiple candidate endpoints in the morning, the target cluster is cluster number 1, and the key point of interest can be inferred to be the company. The specific location of the company can be determined based on cluster number 1. Based on clustering the user's multiple candidate endpoints in the evening, the target cluster is cluster number 2, and the key point of interest can be inferred to be home. The specific location of home can be determined based on cluster number 2.

[0098] For example, a static behavior profile is determined based on the basic static behavior profile and key points of interest. Based on the above example, the static behavior profile shown in Table 3 can be determined.

[0099] Table 3

[0100]

[0101] Among them, Feature 1, Feature 2, ... can be basic static behavioral profiles including: average number of trips per day, differences between weekday and weekend trips, high-frequency departure / arrival time periods, daily activity range, commuting distance, commuting time, average driving speed, and average driving mileage.

[0102] In some embodiments, determining preferred path data based on the frequency characteristics of multiple path segments includes: extracting multiple candidate path segments belonging to a preset time period from the multiple path segments; for each candidate path segment: determining the frequency characteristics of the candidate path segment in the preset time period; if the frequency characteristics of the candidate path segment meet the familiarity condition, using the candidate path segment as the preferred path segment in the preset time period; and determining preferred path data based on the preset time period and the frequency characteristics of the preferred path segments.

[0103] The times of multiple candidate path segments all fall within a preset time period. The preset time period can be set according to actual needs; for example, the preset time period can be the morning peak period, the evening peak period, and the off-peak period.

[0104] Specifically, multiple path segments can be extracted from multiple historical trajectory data using a sliding window, or the multiple historical trajectory data can include all road segments. Based on the time of the multiple path segments, they are divided into candidate path segments belonging to different preset time periods. For example, among the multiple path segments, multiple candidate path segments belonging to the morning rush hour, multiple candidate path segments belonging to the evening rush hour, and multiple candidate path segments belonging to the off-peak hours are determined.

[0105] For each candidate path segment in a preset time period, the frequency characteristics of the path segment are determined based on the multiple candidate path segments in that preset time period. This allows us to determine the frequency characteristics of each candidate path segment. We then determine whether the frequency characteristics of each candidate path segment meet the familiarity criteria. If they do, the candidate path segment is taken as the preferred path segment for that preset time period. Based on the preset time period and the frequency characteristics of the preferred path segments, we determine the preferred path data.

[0106] In some embodiments, frequency features include frequency count and frequency range; determining the frequency features of a candidate path segment within a preset time period includes: obtaining the frequency count of the candidate path segment within the preset time period; determining the frequency range of the candidate path segment within the preset time period based on the frequency count and the number of multiple candidate path segments; and selecting the candidate path segment as a preferred path segment for the preset time period when the frequency features of the candidate path segment meet the familiarity condition, including: selecting the candidate path segment as a preferred path segment for the preset time period when the frequency count of the candidate path segment is greater than a preset frequency count and / or the frequency range of the candidate path segment is greater than a preset frequency range.

[0107] Specifically, for each candidate path segment, the frequency of occurrence of the candidate path segment is counted among multiple candidate path segments in a preset time period; based on the frequency of the candidate path segment and the number of multiple candidate path segments in the preset time period, the frequency of occurrence of the candidate path segment is determined.

[0108] For example, the frequency of candidate path segments appearing in a preset time period is calculated according to formula (6).

[0109] Formula (6): ;

[0110] in, Candidate path segments The frequency of occurrence within the preset time period H is It represents the number of candidate path segments within a preset time period H, where the length of the candidate path segment P is k. It is the frequency of candidate path segment P appearing in a preset time period H.

[0111] Optionally, if the frequency of a candidate path segment is greater than a preset frequency, or if the frequency of a candidate path segment is greater than a preset frequency, then the candidate path segment is used as the preferred path segment for a preset time period.

[0112] Optionally, if the frequency of a candidate path segment is greater than a preset frequency and the relative frequency of the candidate path segment is greater than a preset relative frequency, then the candidate path segment is selected as the preferred path segment for a preset time period. For example, as shown in formula (7).

[0113] Formula (7): ;

[0114] in, It is the frequency of candidate path segment P appearing within a preset time period H. It is a preset frequency. It is the frequency of candidate path segment P appearing within a preset time period H. It is the preset frequency; the values ​​of preset frequency and preset frequency number can be set according to actual needs.

[0115] For example, the preference path data is shown in Table 4.

[0116] Table 4

[0117]

[0118] In the above embodiments, the preferred path segments within the preset time period are determined based on the frequency and frequency of candidate path segments appearing within the preset time period. This allows the user to obtain familiar routes within a specific time period, thereby improving the accuracy of preferred path segments.

[0119] In some embodiments, multidimensional dynamic behavior data includes associated behavior data and forward trajectory data; determining abnormal path data based on the multidimensional dynamic behavior data of multiple path segments includes: for each path segment, processing the associated behavior data of the path segment using a traffic violation determination model to obtain a behavior identifier for the path segment; for each path segment, determining the forward trajectory data of the path segment based on multiple historical trajectory data, processing the forward trajectory data using a path detection model to obtain an anomaly score for the path segment; identifying abnormal path segments among the multiple path segments based on the behavior identifiers and anomaly scores of the multiple path segments; and determining abnormal path data based on the behavior identifiers and anomaly scores of the abnormal path segments.

[0120] Specifically, multiple path segments can be extracted from multiple historical trajectory data using a sliding window, or multiple historical trajectory data can include all road segments.

[0121] For each path segment, obtain the associated behavioral data for that path segment. The associated behavioral data may include: historical driving data and vehicle-road data corresponding to the path segment; historical driving data and vehicle-road data include vehicle speed, seat belt status, traffic light status when passing through intersections, vehicle direction, road direction, driving lane, continuous driving duration, vehicle lane departure status, road curvature, number of lane changes, etc.

[0122] Historical driving data and vehicle-road data corresponding to the route segment are input into the traffic violation determination model. Based on the historical driving data and vehicle-road data, the traffic violation determination model determines whether the user has committed traffic violations such as speeding, running red lights, or driving against traffic while driving on the route segment. The traffic violation determination model outputs behavior identifiers, which can be normal behavior identifiers (indicating no abnormal behavior) or at least one abnormal behavior identifier (indicating the existence of abnormal behavior and determining the type of abnormal behavior). In other words, the traffic violation determination model can determine whether there is abnormal behavior on the route segment, and if so, it can also determine the specific type of abnormality.

[0123] For example, the anomaly type, anomaly identifier, and traffic rule determination description are shown in Table 5.

[0124] Table 5

[0125]

[0126] It should be noted that the traffic violation determination model can be trained by using the associated behavioral data samples of the path segment samples and the behavioral identifier labels of the path segment samples.

[0127] In some embodiments, determining the forward trajectory data of a path segment based on multiple historical trajectory data includes: determining the target trajectory data to which the path segment belongs among the multiple historical trajectory data; and determining the forward trajectory data based on the road segment preceding the path segment in the target trajectory data.

[0128] Specifically, in the target trajectory data to which the path segment belongs, the road segment with the link number preceding the path segment is identified, and the preceding road trajectory data is constructed based on the road segment with the link number preceding the path segment; for example, the path segment is... , representing the link with index i in the target trajectory data, the preceding trajectory data is represented as: .

[0129] The preceding trajectory data is used as a token sequence. The token sequence is input into the path detection model. The path detection model determines the probability distribution of the next road segment given the existence of preceding trajectory data. The predicted probability of the path segment is obtained from the probability distribution. Based on the predicted probability of the path segment, the anomaly score of the path segment is determined.

[0130] In practical applications, the path detection model can be LM-TAD (Language Model based TrajectoryAnomaly Detection), which is built using the ideas of autoregression and masked self-attention model (causal-attention).

[0131] It should be noted that the path detection model is obtained by training the initial path detection model based on the preceding path trajectory data samples, path segment samples, and anomaly scoring labels.

[0132] For example, for a path segment is Anomaly scores can be represented as: ;in, Is it a path segment? The predicted probability.

[0133] After determining the behavior identifier and abnormal score of each path segment, it is determined whether the behavior identifier of the path segment is an abnormal identifier and whether the abnormal score of the path segment is greater than the preset score threshold. If the behavior identifier of the path segment is an abnormal identifier or the abnormal score of the path segment is greater than the preset score threshold, then the path segment is regarded as an abnormal path segment. The preset score threshold can be dynamically adjusted according to actual needs. For example, as shown in formula (8).

[0134] Formula (8): ;

[0135] If the traffic violation determination model outputs an abnormal behavior identifier, then ; This indicates that the abnormal score of the path segment is greater than the preset score threshold.

[0136] For example, if the behavior identifier of a path segment is an abnormal behavior identifier, then the path segment is regarded as an abnormal path segment, and the abnormal path data is determined based on the abnormal path segment and the abnormal behavior identifier, as shown in Table 6.

[0137] Table 6

[0138]

[0139] In Table 6, the abnormal path segments include: the abnormal path segment with ID: BJ-ARTERY-302-W-012. The abnormal behaviors of this abnormal path segment include: speeding, running red lights, and crossing the line.

[0140] In the above embodiments, the traffic violation determination model can quickly determine whether there is abnormal behavior in a path segment, and the path detection model can determine the abnormality score to quantify the abnormality of the path segment. By combining the output results of the traffic violation determination model and the path detection model, abnormal path segments can be identified, avoiding the misjudgment and missed detection problems caused by a single model identifying abnormalities, and improving the accuracy and comprehensiveness of abnormal path segment detection.

[0141] Optionally, the behavior profile generation method includes: determining the navigation path based on the user's behavior profile and navigation demand data.

[0142] For example, the behavior profile generation method can be applied to dynamic behavior profile systems, such as... Figure 3 As shown, the dynamic behavior profiling system includes: a behavior profiling mining unit, a navigation unit, and a behavior profiling database. The behavior profiling mining unit mines key points of interest, preferred path data, and abnormal path data based on the user's historical trajectory data, and determines the behavior profiling based on these data. The behavior profiling is then updated in the behavior profiling database. When a user initiates a navigation request, the navigation unit retrieves the behavior profiling from the database and provides personalized recommendations based on the profiling, thus improving the matching degree between the navigation path output by the navigation unit and the user's behavior.

[0143] The behavior profile mining unit is also used to periodically update the behavior profile based on the user's latest historical trajectory data. Through closed-loop updates, it is ensured that the behavior profile can reflect the user's recent behavior trends in a timely manner, and the latest behavior profile continuously supports applications such as navigation output recommended routes and route optimization.

[0144] The behavioral profile generation method provided in this application determines key interest points (PIs) through spatiotemporal clustering, making them densely distributed trajectory points in both spatial and temporal dimensions, thus improving the accuracy of PI determination. It also determines preference path data through the frequency characteristics of path segments, distinguishing between occasional and long-term preference paths, thereby improving the accuracy of preference path data. Furthermore, it identifies abnormal path data through multi-dimensional dynamic behavioral data of path segments, making the detection of abnormal behavior more comprehensive and effectively detecting abnormal path data across multiple dimensions. This method dynamically mines users' key interest points, preference path data, and abnormal path data from multiple historical trajectory data, improving the real-time nature of behavioral profiles and ensuring that the behavioral profiles match the users' actual behavioral characteristics, thereby enhancing the effectiveness and accuracy of behavioral profiles.

[0145] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0146] Figure 4 A schematic diagram of the behavioral profile generation device provided in this application is shown below. Figure 4 As shown, the behavioral profile generation device 40 provided in this embodiment includes:

[0147] The spatiotemporal clustering module 410 is used to perform spatiotemporal clustering processing based on multiple historical trajectory data of users to obtain key points of interest.

[0148] Extraction module 420 is used to extract multiple path segments from multiple historical trajectory data;

[0149] The preference determination module 430 is used to determine the preferred path data based on the frequency characteristics of multiple path segments;

[0150] The anomaly determination module 440 is used to determine abnormal path data based on the multidimensional dynamic behavior data of multiple path segments;

[0151] The behavior profile determination module 450 is used to determine the user's behavior profile based on key points of interest, preference path data, and abnormal path data.

[0152] In one possible implementation, the preference determination module is used to extract multiple candidate path segments belonging to a preset time period from multiple path segments; for each candidate path segment: determine the frequency characteristics of the candidate path segment in the preset time period; if the frequency characteristics of the candidate path segment meet the familiar route condition, the candidate path segment is used as the preferred path segment in the preset time period; and determine the preferred path data based on the preset time period and the frequency characteristics of the preferred path segment.

[0153] In one possible implementation, the frequency feature includes frequency count and frequency frequency. The preference determination module is used to obtain the frequency count of candidate path segments appearing within a preset time period; determine the frequency frequency of candidate path segments appearing within the preset time period based on the frequency count and the number of multiple candidate path segments; and if the frequency count of a candidate path segment is greater than a preset frequency count, and / or the frequency frequency of a candidate path segment is greater than a preset frequency frequency, the candidate path segment is used as the preferred path segment for the preset time period.

[0154] In one possible implementation, the multidimensional dynamic behavior data includes associated behavior data and forward trajectory data; the anomaly determination module is used to process the associated behavior data of each path segment using a traffic violation judgment model to obtain the behavior identifier of the path segment; for each path segment, based on multiple historical trajectory data, determine the forward trajectory data of the path segment, and process the forward trajectory data using a path detection model to obtain the anomaly score of the path segment; based on the behavior identifiers and anomaly scores of multiple path segments, identify the abnormal path segment among multiple path segments; and based on the behavior identifiers and anomaly scores of the abnormal path segment, determine the abnormal path data.

[0155] In one possible implementation, the multidimensional dynamic behavior data includes associated behavior data and forward trajectory data; the anomaly determination module is used to determine the target trajectory data to which the path segment belongs among multiple historical trajectory data; and to determine the forward trajectory data based on the road segment preceding the path segment in the target trajectory data.

[0156] In one possible implementation, an extraction module is used to select path segments from the sequence of road segments included in the historical trajectory data for each historical trajectory data point using a sliding window.

[0157] In one possible implementation, a spatiotemporal clustering module is used to extract the start and end points from multiple historical trajectory data of the user to obtain multiple candidate points of interest; determine the spatiotemporal weights of the multiple candidate points of interest based on their time and location; perform spatiotemporal clustering processing on the multiple candidate points of interest to obtain multiple clusters; select a target cluster from the multiple clusters based on the spatiotemporal weights of the candidate points of interest included in the multiple clusters; and determine key points of interest based on the location of the candidate points of interest included in the target cluster.

[0158] In one possible implementation, the spatiotemporal clustering module is used to determine the neighboring interest points of each candidate interest point among multiple candidate interest points; determine the temporal weight based on the preset time and the time of the candidate interest points; determine the spatial weight based on the location of the neighboring interest points and the location of the candidate interest points; and determine the spatiotemporal weight of the candidate interest points based on the temporal weight and the spatial weight.

[0159] The behavior profile generation device provided in this embodiment can execute the behavior profile generation method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0160] The behavior profile generation device provided in this application determines key interest points through spatiotemporal clustering, making these key interest points densely distributed trajectory points in both spatial and temporal dimensions, thus improving the accuracy of key interest point determination. It also determines preferred path data through the frequency characteristics of path segments, distinguishing between occasional and long-term preferred paths, thereby improving the accuracy of preferred path data. Furthermore, it determines abnormal path data through multi-dimensional dynamic behavior data of path segments, making the detection of abnormal behavior more comprehensive and effectively detecting abnormal path data across multiple dimensions. This method dynamically mines key interest points, preferred path data, and abnormal path data of users through multiple historical trajectory data, improving the real-time performance of the behavior profile, ensuring that the behavior profile matches the user's actual behavioral characteristics, and enhancing the effectiveness and accuracy of the behavior profile.

[0161] Figure 5 A schematic diagram of the structure of the electronic device provided in this application. Figure 5 As shown, the electronic device 50 provided in this embodiment includes at least one processor 501 and a memory 502. Optionally, the device 50 further includes a communication component 503. The processor 501, memory 502, and communication component 503 are connected via a bus.

[0162] In a specific implementation, at least one processor 501 executes computer execution instructions stored in memory 502, causing at least one processor 501 to perform the above-described method.

[0163] The specific implementation process of processor 501 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0164] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0165] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0166] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0167] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0168] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0169] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0170] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0171] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0172] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0173] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0174] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0175] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0176] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for generating behavioral profiles, characterized in that, include: Spatiotemporal clustering is performed on multiple historical trajectory data of users to obtain key points of interest; Extract multiple path segments from the multiple historical trajectory data; Based on the frequency characteristics of the multiple path segments, the preferred path data is determined; Based on the multidimensional dynamic behavior data of the multiple path segments, abnormal path data is determined; The user's behavioral profile is determined based on the key points of interest, the preference path data, and the abnormal path data.

2. The method according to claim 1, characterized in that, The step of determining preferred path data based on the frequency characteristics of the multiple path segments includes: From the multiple path segments, extract multiple candidate path segments belonging to a preset time period; For each candidate path segment: determine the frequency characteristics of the candidate path segment in the preset time period; if the frequency characteristics of the candidate path segment meet the familiarity condition, the candidate path segment is taken as the preferred path segment in the preset time period. Based on the preset time period and the frequency characteristics of the preferred path segment, the preferred path data is determined.

3. The method according to claim 2, characterized in that, The frequency features include frequency count and frequency frequency; determining the frequency features of the candidate path segment within the preset time period includes: Obtain the frequency of occurrence of the candidate path segment within the preset time period; The frequency of the candidate path segment in the preset time period is determined based on the frequency and the number of the plurality of candidate path segments; The step of selecting a candidate path segment as the preferred path segment for the preset time period when the frequency feature of the candidate path segment satisfies the familiarity condition includes: If the frequency of the candidate path segment is greater than a preset frequency, and / or the relative frequency of the candidate path segment is greater than a preset relative frequency, the candidate path segment will be used as the preferred path segment for the preset time period.

4. The method according to claim 1, characterized in that, The multidimensional dynamic behavior data includes associated behavior data and forward trajectory data; The step of determining abnormal path data based on the multidimensional dynamic behavior data of the multiple path segments includes: For each path segment, the associated behavioral data of the path segment is processed by a traffic violation determination model to obtain the behavioral identifier of the path segment; For each path segment, based on the multiple historical trajectory data, the preceding trajectory data of the path segment is determined, and the preceding trajectory data is processed by the path detection model to obtain the anomaly score of the path segment; Based on the behavior identifiers and anomaly scores of the plurality of path segments, an abnormal path segment is identified among the plurality of path segments; Abnormal path data is determined based on the behavior identifier and the abnormal score of the abnormal path segment.

5. The method according to claim 4, characterized in that, The step of determining the forward trajectory data of the path segment based on the multiple historical trajectory data includes: Among the multiple historical trajectory data, determine the target trajectory data to which the path segment belongs; Based on the target trajectory data, the road segment preceding the path segment is determined to obtain the preceding trajectory data.

6. The method according to any one of claims 1 to 5, characterized in that, The extraction of multiple path segments from the multiple historical trajectory data includes: For each historical trajectory data, a path segment is selected from the road segment sequence included in the historical trajectory data using a sliding window.

7. The method according to any one of claims 1 to 5, characterized in that, The process of spatiotemporal clustering based on multiple historical trajectory data of users to obtain key points of interest includes: Extract the start and end points from multiple historical trajectory data of the user to obtain multiple candidate points of interest; The spatiotemporal weights of the multiple candidate interest points are determined based on their time and location. Spatiotemporal clustering is performed on the multiple candidate points of interest to obtain multiple clusters; Based on the spatiotemporal weights of the candidate interest points included in the plurality of clusters, a target cluster is selected from the plurality of clusters; Key points of interest are determined based on the locations of candidate points of interest included in the target cluster.

8. The method according to claim 7, characterized in that, The step of determining the spatiotemporal weights of the multiple candidate interest points based on their time and location includes: For each candidate point of interest, determine the neighboring points of interest among the plurality of candidate points of interest; The time weight is determined based on the preset time and the time of the candidate points of interest; Spatial weights are determined based on the locations of the neighboring interest points and the candidate interest points; The spatiotemporal weights of the candidate points of interest are determined based on the time weights and the spatial weights.

9. A behavioral profile generation device, characterized in that, The device includes: The spatiotemporal clustering module is used to perform spatiotemporal clustering based on multiple historical trajectory data of users to obtain key points of interest; The extraction module is used to extract multiple path segments from the multiple historical trajectory data; The preference determination module is used to determine the preferred path data based on the frequency characteristics of the multiple path segments; The anomaly determination module is used to determine abnormal path data based on the multidimensional dynamic behavior data of the multiple path segments; The behavior profile determination module is used to determine the user's behavior profile based on the key points of interest, the preference path data, and the abnormal path data.

10. An electronic device / computer-readable storage medium / computer program product, characterized in that, The electronic device includes: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method as described in any one of claims 1 to 8; The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 8; The computer program product includes computer execution instructions that, when executed by a processor, implement the method as described in any one of claims 1 to 8.

Images