A method for detecting user linkage behavior patterns in a spatiotemporal social network

Abstract

The application designs a user linkage behavior pattern detection method in a space-time social network; for a given space-time social network and a query time period, when processing linkage behavior pattern detection, according to the check-in behavior of a user, an active user set in a T period and a corresponding visited, i.e. check-in, interest point set S are selected T The interest point set is the activity place of the linkage behavior pattern in the period; for a given spatial distance r, a range query is performed in the interest point S T , interest point pairs with a spatial layer distance less than or equal to r are found; users who have visited the interest point pairs are selected to form linkage behavior pattern candidates; the social correlation between the users is verified to obtain the final detected user linkage behavior pattern; the linkage behavior pattern is detected from the perspective of a space-time social network, the social topological structure and the behavior activity (check-in) of a user in a geographical space layer are fully combined, and the query processing system in the computer data management field is enriched and expanded.

Description

Technical Field

[0001] This invention belongs to the field of spatiotemporal social network technology, and in particular relates to a method for detecting user interaction behavior patterns in spatiotemporal social networks. Background Technology

[0002] With the popularization of online social networks and the rapid development of diverse positioning technologies, a new type of social network, the Spatio-Temporal Social Network (STSN), has emerged, exemplified by platforms such as Dianping, Sina Weibo, Foursquare, and Yelp. These platforms contain complex and rich user-generated data, such as user personal information, check-in data, and review data, bringing new opportunities for analyzing user mobility behavior. User interaction behavior refers to the highly overlapping activity trajectories of multiple users within the same time period. Detecting patterns of user interaction behavior is an important means of analyzing user group mobility and has wide applications in user group behavior analysis, precision marketing, and personalized recommendations.

[0003] The data model for spatiotemporal social networks is attached. Figure 1 As shown in (a), this is a two-layer network structure, mainly consisting of a social layer and a geospatial layer. Specifically, as shown in the attached diagram... Figure 1 (a) The social layer is often modeled as an undirected graph G = (V, E), where the set of nodes V represents social users, and any two users connected by an edge e in the set of edges E have a social relationship (such as a friend relationship). In the geospatial layer, there are semantically labeled locations (called points of interest), such as cafes and libraries. Check-in is one of the most common user behaviors on spatiotemporal social networking platforms, recording the offline geospatial activities of social network users. (See attached image) Figure 1 As shown in (b), each check-in record consists of a 3-tuple (u, s, t), representing that user u visited point of interest s at time t. The rich data types inherent in spatiotemporal social networks bring new opportunities for the detection of linked behavior patterns. The definition of user linked behavior in spatiotemporal social networks is: the behavior of multiple users visiting the same or adjacent points of interest within the same time period, where these users have relatively close social relationships. In a given time period, the tuple consisting of users who exhibited linked behavior and their corresponding interests is called a user linked behavior pattern.

[0004] In the area of ​​user mobility analysis, Chinese patent "CN113990075A A method and system for traffic flow allocation by fusing traffic survey data and trajectory data" provides a method for detecting and allocating traffic flow. This patent first preprocesses the urban road network by encoding road segments based on the network's topology. Second, it matches user trajectory points to corresponding road segments, forming user travel trajectories and constructing trajectory sets. Then, it queries the correlation between given survey locations and trajectories, calculates trajectory expansion coefficients, and updates the remaining traffic flow of road segments based on the expansion coefficient values ​​of the aggregated survey locations. Finally, it aggregates the trajectory coefficients onto road segments to obtain the corresponding traffic flow, thereby achieving traffic flow monitoring and dynamic allocation. The technical solution described in Chinese patent "CN113990075A A method and system for traffic flow allocation by fusing traffic survey data and trajectory data" primarily analyzes user mobility in the road traffic network by fusing user trajectory data into road segments. However, in spatiotemporal social networks, there is no user trajectory data. Furthermore, in real life, users tend to be more concerned about the privacy of their personal trajectory data and will not choose to disclose their movement trajectory data. Therefore, the technology of this patent cannot be applied to spatiotemporal social network scenarios.

[0005] Chinese patent "CN114119840A A Method for Generating Heatmaps of Massive Trajectory Data" provides a method for producing heatmaps of vehicle trajectory data. This patent primarily uses spatial aggregation technology to aggregate massive directional trajectory lines into directional heatmap data, and then uses a dynamic rendering method for the directional heatmap dataset to generate heatmaps of vehicle trajectories, providing a powerful tool for visualizing and analyzing urban vehicle and user mobility. While this patented technology can macroscopically present user mobility, it cannot capture the interactive behavioral patterns between different individual users; that is, it cannot detect highly overlapping activity trajectories of different users within a continuous time period. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention proposes a method for detecting user interaction behavior patterns in spatiotemporal social networks.

[0007] For a given spatiotemporal social network and a query time period, when processing the detection of linked behavior patterns, the first step is to select the set of active users within the time period T based on the user's check-in behavior. and the corresponding set S of points of interest that are visited (i.e., checked in). T The set of points of interest represents the activity locations of linked behavioral patterns within that time period; secondly, for a given spatial distance r, at point of interest S T Perform a range query to find pairs of interest points whose spatial distance is less than or equal to r; then, from... Users who have visited the point of interest pair are selected to form candidate linkage behavior patterns; finally, the social correlation between these users is verified to obtain the final detected user linkage behavior patterns.

[0008] A method for detecting user interaction behavior patterns in a spatiotemporal social network includes the following steps:

[0009] Step 1: Obtain the activity locations of users who exhibited linked behaviors;

[0010] In spatiotemporal social networks, check-in behavior is represented by triples, denoted as c, where c = (v, s, t), indicating that social network user v visited point of interest s at time t. To detect interactive behavior patterns in spatiotemporal social networks, check-in records C within a given time period T are selected. T ={c|ct∈T}, these users who checked in are active users during the time period T. And check-in record C T The set of points of interest in the data is denoted as S. T The spatiotemporal social network includes a social network G = (V, E), a set of geospatial interest points S, and historical check-in records of social users in the geospatial layer. Each check-in record c consists of a 3-tuple c = (v, s, t).

[0011] Step 2: By filtering the spatiotemporal social network's geospatial layer, a set of users constituting the linked behavioral pattern is obtained;

[0012] For any two social users u and v, if there is a linked behavior pattern in time period T, then the two users have visited the same point of interest in that time period, or the distance between the different points of interest they visited does not exceed r in the geographic space; if the distance between the two points of interest is greater than r, then the users who visited the two points of interest in the same time period do not belong to linked behavior.

[0013] Based on the above, filtering is performed at the geospatial level; the set S obtained in step 1 is calculated. T Any two points of interest s i s j The distance between D(s) i s j If D(s) i s j If ) > r, then during time period T, the number of visits to point of interest s is... i User Vi and Access Points of Interest j User V j There is no linked behavior pattern; conversely, set V i With V j Users within the system may exhibit interconnected behavioral patterns.

[0014] On the other hand, users accessing the same point of interest may engage in collaborative behavior, that is, regarding point of interest s... i or s i Users form a collaborative behavior pattern;

[0015] In the above case, by filtering the geospatial layer, a set of users constituting the linkage behavior pattern is obtained, which is called the candidate set.

[0016] Step 3: By performing social relevance filtering on the set of users with linked behaviors obtained in Step 2, a new set of user candidates constituting the linked behavior pattern is obtained.

[0017] Although a set of potential users constituting a linkage behavior pattern was obtained in step 2, any combination of users in the set constitutes a linkage behavior pattern; therefore, the combinations formed by the set of potential users are filtered from the social layer.

[0018] First, social relevance is defined as:

[0019] Definition 1, Social Relevance: Let the social layer in a spatiotemporal social network be G = (V, E), where the set of nodes V represents social users, and any two users connected by an edge e in the set of edges E have a social relationship, such as a friendship. Let path(u, v) represent a social path between nodes u and v, whose length is the number of edges traversed by the path, and the social distance Dist(u, v) between nodes u and v be the length of the shortest social path between these two nodes. Then, the social relevance SR(u, v) between these users can be expressed as:

[0020] SR(u,v)=max(0,1-Dist(u,v) / maxDist) (1)

[0021] Where maxDist represents the maximum social distance that the linked behavior pattern query can tolerate;

[0022] For a set of users who visit the same or nearby points of interest at the same time, if they form a linked behavior pattern, then the users in this set have high social relevance. In real life, if two users have a linked behavior pattern, then these two users usually have a close social relationship, that is, they have a small social distance. In addition, based on the small-world phenomenon, the four-degree segmentation theory, and the three-degree connection phenomenon, the tolerance value of the maximum social distance, maxDist, is set to a small constant. Assuming maxDist = 3, when the social distance between two users is greater than 3, that is, Dist(u, v) is greater than 3, then (1-Dist(u, v) / maxDist)) < 0, then SR(u, v) = 0. Therefore, these two users cannot form a linked behavior pattern at this time.

[0023] Definition 2, Lower bound of social distance between sets: two sets of users V i V j The lower bound of social distancing between people: LowD(V) i V j Let be the minimum social distance between nodes in these two sets, therefore:

[0024]

[0025] Where Vi is the set of users who have visited point of interest i;

[0026] In user set V, nodes connected by social edges to users in other user sets are called boundary nodes V.cro, while nodes connected only by social edges to users within their own set are called internal nodes V.int, and V = V.cro ∪ V.int; the social distance between two sets is the minimum distance between their corresponding boundary nodes, therefore:

[0027]

[0028] By calculating two potential user sets V i V j The lower bound of the distance between them reduces the search space; if LowD(V) i V j If LowD(V) > maxDist, then users in these two sets cannot form a linked behavior pattern; conversely, if LowD(V) > maxDist, then users in these two sets cannot form a linked behavior pattern. i V j If ) ≤ maxDist, then the users in these two sets constitute a linked behavior pattern, and candidate linked behavior pattern P is generated. * =({V i V j}, {s i s j}, T) as the input for the next step; the candidate P of this linkage behavior pattern * Indicates the point of interest s to be visited i User V i With access to points of interest j User V j A collaborative behavior pattern is formed between them; in addition, access to points of interest s i or s j Users also form interactive behavior patterns, so the candidates for interactive behavior patterns also include (V) i s i (T) and (V) j s j The candidate linkage behavior patterns obtained in this step (T) are all used as inputs for the next step.

[0029] Step 4: Verify the candidate results of the linkage behavior patterns obtained in Step 3;

[0030] In the candidate linked behavior patterns obtained in step 3, each user is a set, indicating that users in this set may form linked behavior patterns. Verification is performed on the specific linked behavior patterns in each candidate; there are two cases:

[0031] (1) Users with linked behavior patterns originate from a single set of candidate users;

[0032] When users exhibiting a linked behavior pattern originate from a single candidate user set, then the participating users in this type of linked behavior pattern access the same point of interest. Taking Vi as an example, this means that all participating users in the linked behavior pattern belong to Vi, and within time period T, all users in this set access s. i For V i For any two users u and v, there exists a candidate pattern of linked behavior p = ({u, v}, s) i , T), calculate the length of the shortest social path between u and v, i.e., the social distance Dist(u, v). If Dist(u, v) ≤ maxDist, then set p = ({u, v}, s i ,T) are added to the final set of linked behavior patterns P;

[0033] (2) Users with linked behavior patterns originate from different sets of candidate users;

[0034] When users exhibiting linked behavioral patterns originate from different candidate user sets V i V j At that time, assume there is v x ∈V i v y ∈V j So, during time period T, v x visited s i v y visited s j The candidate for the linked behavior pattern is p = ({v x v y}, {s i s j}, T); will be calculated by v x With v y Social distancing Dist(v x v y To verify whether the candidate p is a linked behavior pattern: If Dist(v) x v y If )≤maxDist, then p=({v x vy}, {s i s j}, T) are added to the final set of linked behavior patterns P;

[0035] The linked behavior pattern satisfies the transitivity principle, that is, in time period T, if u and v have a linked behavior pattern, and v and w have a linked behavior pattern, then u and w have a linked behavior pattern, and the participating users in this linked behavior pattern are {u, v, w}. According to the transitivity principle of linked behavior patterns, not only linked behavior patterns formed by two users can be detected, but also linked behavior patterns formed by multiple users can be detected.

[0036] Beneficial technical effects of the present invention:

[0037] Compared with existing technologies, this invention detects linked behavior patterns from the perspective of spatiotemporal social networks, fully integrating social topology and user behavior (check-in) at the geospatial layer, enriching and expanding the query processing system in the field of computer data management. The technical solution proposed in this invention introduces a spatial distance-based filtering technique, supplementing the detection of linked behavior patterns that only consider visits to the same point of interest. The proposed pruning strategy based on social relevance transforms social relevance into the calculation of social distance, using the lower bound of the social distance of user sets to reduce the search space, avoiding combined searches between different user sets, thereby improving search efficiency. Based on the transitivity of linked behavior patterns, the technical solution provided by this invention can detect not only linked behavior patterns formed by two users, but also linked behavior patterns formed by multiple users. Attached Figure Description

[0038] Figure 1 An example diagram of a spatiotemporal social network data model according to an embodiment of the present invention, wherein Figure a shows the hierarchical structure of the spatiotemporal social network; Figure b shows the check-in information;

[0039] Figure 2 A flowchart of a method for detecting user interaction behavior patterns in a spatiotemporal social network according to an embodiment of the present invention. Detailed Implementation

[0040] The present invention will be further described below with reference to the accompanying drawings and embodiments;

[0041] The purpose of this invention is to provide a method for detecting user interaction behavior patterns in spatiotemporal social networks. Specifically, this invention utilizes user activity data from spatiotemporal social networks, including check-in data and social relationships between users, and employs a "filter-verify" processing framework based on social topology to analyze and process social user interaction behavior patterns under input time conditions. Specifically, in the filtering stage, pruning is performed based on spatial distance. For users accessing different points of interest, it is determined whether interaction behavior is possible, and then, based on social distance, user groups with potential interaction behavior patterns are selected as candidate result sets. In the verification stage, verification is performed by calculating the social correlation between users in the candidate result sets, thereby realizing the detection of user interaction behavior patterns in spatiotemporal social networks.

[0042] A spatiotemporal social network mainly consists of a social network G = (V, E), a set of geospatial points of interest S, and historical check-in records of social users at the geospatial layer. Each check-in record c consists of a tuple c = (v, s, t). This invention aims to return user groups with highly overlapping offline activities in the spatiotemporal social network. That is, each returned user group has high social relevance at the social layer and their activity ranges at the geospatial layer are relatively close within a given query time period. An example diagram of the spatiotemporal social network data model is attached. Figure 1 As shown;

[0043] For a given spatiotemporal social network and a query time period, when processing the detection of linked behavior patterns, the first step is to select the set of active users within the time period T based on the user's check-in behavior. and the corresponding set S of points of interest that are visited (i.e., checked in). T The set of points of interest represents the activity locations of linked behavioral patterns within that time period; secondly, for a given spatial distance r, at point of interest S T Perform a range query to find pairs of interest points whose spatial distance is less than or equal to r; then, from... Users who have visited the point of interest pair are selected to form candidate linkage behavior patterns; finally, the social correlation between these users is verified to obtain the final detected user linkage behavior patterns.

[0044] A method for detecting user interaction behavior patterns in spatiotemporal social networks, the process of which is attached. Figure 2 As shown, it includes the following steps:

[0045] Step 1: Obtain the activity locations of users who exhibited linked behaviors;

[0046] In spatiotemporal social networks, check-in is a common behavior, recording the visits of social users to points of interest in the spatial layer. This is usually represented by triples. Let c be a check-in record, then c = (v, s, t), indicating that social network user v visited point of interest s at time t. To detect interactive behavior patterns in spatiotemporal social networks, check-in records C within a given time period T are selected. T ={c|ct∈T}, these users who checked in are active users during the time period T. And check-in record C T The set of points of interest in the data is denoted as S. T The spatiotemporal social network mainly includes a social network G = (V, E), a set of geospatial interest points S, and historical check-in records of social users in the geospatial layer. Each check-in record c consists of a 3-tuple c = (v, s, t).

[0047] Step 2: By filtering the spatiotemporal social network's geospatial layer, a set of users constituting the linked behavioral pattern is obtained;

[0048] For any two social users u and v, if there is a linked behavior pattern in time period T, then the two users have visited the same point of interest in that time period, or the distance between the different points of interest they visited does not exceed r (r is usually a very small positive number, meaning that the two locations are very close); if the distance between the two points of interest is greater than r, then the users who visited these two points of interest in the same time period do not belong to linked behavior.

[0049] Based on the above, filtering is performed at the geospatial level; the set S obtained in step 1 is calculated. T Any two points of interest s i s j The distance between D(s) i s j If D(s) i s j If ) > r, then during time period T, the number of visits to point of interest s is... i User Vi and Access Points of Interest j User V j There is no linked behavior pattern; conversely, set V i With V j Users within the system may exhibit interconnected behavioral patterns.

[0050] On the other hand, users accessing the same point of interest may engage in collaborative behavior, that is, regarding point of interest s... i or s j Users may form collaborative behavior patterns;

[0051] In the above situation, by filtering the geospatial layer, a set of users that may constitute a linkage behavior pattern is obtained, which is called the candidate set.

[0052] Step 3: By performing social relevance filtering on the set of users with linked behaviors obtained in Step 2, a new set of user candidates constituting the linked behavior pattern is obtained.

[0053] Although a set of potential users constituting a linkage behavior pattern was obtained in step 2, any number of user combinations in the set may constitute a linkage behavior pattern; if each combination is verified, it will increase the amount of computation and lead to inefficiency. Therefore, the combinations formed by the set of potential users are filtered from the social layer.

[0054] First, social relevance is defined as:

[0055] Definition 1, Social Relevance: Let the social layer in a spatiotemporal social network be G = (V, E), where the set of nodes V represents social users, and any two users connected by an edge e in the set of edges E have a social relationship, such as a friendship. Let path(u, v) represent a social path between nodes u and v, whose length is the number of edges traversed by the path, and the social distance Dist(u, v) between nodes u and v be the length of the shortest social path between these two nodes. Then, the social relevance SR(u, v) between these users can be expressed as:

[0056] SR(u,v)=max(0,1-Dist(u,v) / maxDist) (1)

[0057] Where maxDist represents the maximum social distance that the linked behavior pattern query can tolerate;

[0058] For a set of users who visit the same or nearby points of interest at the same time, if they form a linked behavior pattern, then the users in this set have high social relevance. In real life, if two users have a linked behavior pattern, then these two users usually have a close social relationship, that is, they have a small social distance. In addition, based on the small-world phenomenon, the four-degree segmentation theory, and the three-degree connection phenomenon, the tolerance value of the maximum social distance, maxDist, is set to a small constant. Assuming maxDist = 3, when the social distance between two users is greater than 3, that is, Dist(u, v) is greater than 3, then (1-Dist(u, v) / maxDist)) < 0, then SR(u, v) = 0. Therefore, these two users cannot form a linked behavior pattern at this time.

[0059] Definition 2, Lower bound of social distance between sets: two sets of users V i V jThe lower bound of social distancing between people: LowD(V) i V j Let be the minimum social distance between nodes in these two sets, therefore:

[0060]

[0061] Where Vi is the set of users who have visited point of interest i;

[0062] In user set V, nodes connected by social edges to users in other user sets are called boundary nodes V.cro, while nodes connected only by social edges to users within their own set are called internal nodes V.int, and V = V.cro ∪ V.int; the social distance between two sets is the minimum distance between their corresponding boundary nodes, therefore:

[0063]

[0064] By calculating two potential user sets V i V j The lower bound of the distance between them reduces the search space; if LowD(V) i V j If LowD(V) > maxDist, then users in these two sets cannot form a linked behavior pattern; conversely, if LowD(V) > maxDist, then users in these two sets cannot form a linked behavior pattern. i V j If ) ≤ maxDist, then the users in these two sets may constitute a linked behavior pattern. Therefore, candidate linked behavior pattern P is generated. * =({V i V j}, {s i s j}, T) as the input for the next step; the candidate P of this linkage behavior pattern * Indicates the point of interest s to be visited i User V i With access to points of interest j User V j There may be interactive behavior patterns between them; in addition, access to points of interest s i or s j Users may also form interactive behavior patterns, so candidates for interactive behavior patterns also include (V) i s i (T) and (V) j s j The candidate linkage behavior patterns obtained in this step (T) are all used as inputs for the next step.

[0065] Step 4: Verify the candidate results of the linkage behavior patterns obtained in Step 3;

[0066] In the candidate linked behavior patterns obtained in step 3, each user is a set, indicating that users in this set may form linked behavior patterns. Verification is performed on the specific linked behavior patterns in each candidate; there are two cases:

[0067] (1) Users with linked behavior patterns originate from a single set of candidate users;

[0068] When users exhibiting interconnected behavior patterns originate from a single set of candidate users, then the participating users in this type of interconnected behavior pattern access the same point of interest; without loss of generality, let V i Taking an example, all users participating in the linked behavior pattern belong to V. i Furthermore, during time period T, all users in this set accessed s. i For V i For any two users u and v, there exists a candidate pattern of linked behavior p = ({u, v}, s) i , T), calculate the length of the shortest social path between u and v, i.e., the social distance Dist(u, v). If Dist(u, v) ≤ maxDist, then set p = ({u, v}, s i ,T) are added to the final set of linked behavior patterns P;

[0069] (2) Users with linked behavior patterns originate from different sets of candidate users;

[0070] When users exhibiting linked behavioral patterns originate from different candidate user sets V i V j At that time, assume there is v x ∈V i v y ∈V j So, during time period T, v x visited s i v y visited s j The candidate for the linked behavior pattern is p = ({v x v y}, {s i s j}, T); will be calculated by v x With v y Social distancing Dist(v x v y To verify whether the candidate p is a linked behavior pattern: If Dist(v) x v y If )≤maxDist, then p=({v x v y}, {s i sj}, T) are added to the final set of linked behavior patterns P;

[0071] The linked behavior pattern satisfies the transitivity principle, that is, in time period T, if u and v have a linked behavior pattern, and v and w have a linked behavior pattern, then u and w have a linked behavior pattern, and the participating users in this linked behavior pattern are {u, v, w}. According to the transitivity principle of linked behavior patterns, not only linked behavior patterns formed by two users can be detected, but also linked behavior patterns formed by multiple users can be detected.

[0072] This invention aims to provide a method for detecting collaborative behavior patterns in spatiotemporal social networks. It proposes a "filter-verify" processing framework based on social topology to analyze and process social user collaborative behavior patterns based on input time conditions. In the filtering stage, for a given query time period, the distance between the points of interest visited by different users is used to determine whether collaborative behavior exists between users visiting different points of interest, thus supplementing the detection of collaborative behavior only between users visiting a single point of interest. Furthermore, a pruning strategy based on the lower bound of the social distance between user sets is proposed. If the lower bound of the social distance between two user sets is greater than the social distance tolerance value, then there is no collaborative behavior between the users in these two sets. This strategy avoids combined searches between different user sets, reduces the understanding space, and thus improves search efficiency.

Claims

1. A method for detecting user interaction behavior patterns in a spatiotemporal social network, characterized in that, Includes the following steps: Step 1: Obtain the activity locations of users who exhibited linked behaviors; Step 2: By filtering the spatiotemporal social network's geospatial layer, a set of users constituting the linked behavioral pattern is obtained; For any two social users and If within a time period If a linked behavior pattern exists, then these two users either visited the same point of interest during that time period, or the geographical distance between the different points of interest they visited did not exceed [a certain value]. ; If the distance between two points of interest is greater than Therefore, users who visit these two points of interest at the same time are not considered to be engaging in linked behavior. Based on the above, filtering is performed at the geospatial level; the set obtained in step 1 is calculated. Any two points of interest , Distance between ,if So in During the time period, visit points of interest users Access Points of Interest users There is no linked behavior pattern; conversely, a set Users within the group may form interactive behavior patterns; On the other hand, users accessing the same point of interest may engage in collaborative behavior, that is, regarding the point of interest or... Users form a collaborative behavior pattern; In the above case, by filtering the geospatial layer, a set of users constituting the linkage behavior pattern is obtained, which is called the candidate set. Step 3: By performing social relevance filtering on the set of users with linked behaviors obtained in Step 2, a new set of user candidates constituting the linked behavior pattern is obtained. Step 4: Verify the candidate results of the linkage behavior patterns obtained in Step 3.

2. The method for detecting user interaction behavior patterns in a spatiotemporal social network according to claim 1, characterized in that, Step 1, which involves obtaining the activity locations of users exhibiting linked behaviors, specifically refers to: In spatiotemporal social networks, triples are used to represent check-in behavior, recording... For a single sign-in record, then , indicating social network users exist Points of interest were visited frequently. To detect collaborative behavior patterns in spatiotemporal social networks, a given time period was selected. Sign-in records These users who checked in were for that time period. active users and check-in records The set of points of interest in the middle is denoted as .

3. The method for detecting user interaction behavior patterns in a spatiotemporal social network according to claim 2, characterized in that, The spatiotemporal social network includes social networks. Geospatial layer set of points of interest And the historical check-in records of social users at the geospatial layer, each check-in record Composed of 3-tuples .

4. The method for detecting user interaction behavior patterns in a spatiotemporal social network according to claim 1, characterized in that, Step 3 specifically involves: Although a set of potential users constituting a linkage behavior pattern was obtained in step 2, any combination of users in the set constitutes a linkage behavior pattern; therefore, the combinations formed by the set of potential users are filtered from the social layer. First, social relevance is defined as: Definition 1, Social Relevance: Let the social layer in a spatiotemporal social network be... , where the set of nodes Representing social users, side collection any edge in The two connected users have a social relationship, such as a friend relationship; (Note:) Represents a node and A social path between nodes, the length of which is the number of edges traversed by the path, and the number of nodes. and social distancing Given the length of the shortest social path between these two nodes, the social relevance between these users is... It can be represented as: （1） in, This indicates the maximum social distance that can be tolerated for linked behavior patterns; For a set of users who visit the same or nearby points of interest at the same time, if they exhibit a linked behavior pattern, then the users in this set have high social relevance. In real life, if two users exhibit a linked behavior pattern, then these two users usually have a close social relationship, i.e., a small social distance. Furthermore, based on the small-world phenomenon, the four-degrees-of-separation theory, and the three-degrees-of-connection phenomenon, the tolerance value for the maximum social distance is... Let it be a small constant; assume When the social distance between two users is greater than 3, that is... If it is greater than 3, then ,but Therefore, these two users could not form a coordinated behavioral pattern during that period. Definition 2, Lower bound of social distance between sets: two sets of users , The lower bound of social distancing The minimum social distance between nodes in these two sets is given by: （2） Where Vi is the set of users who have visited point of interest i; User set Nodes that have social edges connecting users in another user set are called boundary nodes. Nodes that are only connected to other users within their own set are internal nodes. ,and The social distance between two sets is the minimum distance between their corresponding boundary nodes, therefore: （3） By calculating two sets of potential users , The lower bound of the distance between them reduces the search space. If the users in these two sets cannot form a linked behavior pattern, then the users in these two sets cannot form a linked behavior pattern; conversely, if If the users in these two sets constitute a linked behavior pattern, then candidate linked behavior patterns are generated. As input for the next step; this linked behavior pattern candidate Indicates access to points of interest users Access Points of Interest users In addition, there is a pattern of interconnected behavior; and access to points of interest. or Users also form interactive behavior patterns, so other candidate interactive behavior patterns include... and The candidate linkage behavior patterns obtained in this step are all used as input for the next step.

5. The method for detecting user interaction behavior patterns in a spatiotemporal social network according to claim 1, characterized in that, Step 4: In the candidate linked behavior patterns obtained in Step 3, each user is a set, indicating that users in the set form a linked behavior pattern. Verify the specific linked behavior pattern in each candidate.

6. The method for detecting user interaction behavior patterns in a spatiotemporal social network according to claim 5, characterized in that, For each candidate, the specific linkage behavior pattern is verified, and there are two cases: (1) Users with linked behavior patterns originate from a single set of candidate users; When users of a linked behavior pattern originate from a single set of candidate users, then the participating users of this type of linked behavior pattern visit the same point of interest. by Taking the example of a linked behavior pattern, all participating users belong to... And in During the time period, all users in this set accessed ;for Any two users , There are candidate collaborative behavior patterns. ,calculate and The length of the shortest social path between them, i.e., social distance ,if Then Added to the final set of linked behavior patterns ; (2) Users exhibiting linked behavior patterns originate from different sets of candidate users; When users exhibiting linked behavioral patterns originate from different sets of candidate users , At that time, assuming there is , So, in During the period, Visited , Visited The candidate for the coordinated behavior pattern is ; will be calculated and social distancing To verify the candidate Is it a linked behavior pattern: If Then Added to the final set of linked behavior patterns ; The interconnected behavioral pattern satisfies the principle of transitivity, that is, in During the period, if and There is a pattern of interconnected behavior. and If there is a pattern of interconnected behaviors, then and There is a collaborative behavior pattern, and the users participating in this collaborative behavior pattern are... Based on the principle of the transitivity of linked behavior patterns, it can detect not only linked behavior patterns formed by two users, but also linked behavior patterns formed by multiple users.

Images