A Geographic Community Search Method Based on Time Windows

Abstract

The present invention belongs to the field of computer applications and discloses a method for retrieving geographical communities based on time windows, which includes the following steps: Step 1, preprocess the geographical social network data set with time attributes; Step 2, obtain all communities in different time periods, and then perform k-core decomposition on the communities belonging to different time periods; Step 3, establish a TDC-index index table structure; Step 4, according to the input, use the index structure to query and screen out the geographical communities that meet the distance threshold; Step 5, return the query result. The present invention fully explores the correlation between time windows and user geographical locations, making the queried geographical communities more time-effective, and pre-establishes an efficient index structure for users within a certain time window and geographical range, providing an efficient query of geographical communities when different time windows, tightness, and distance thresholds are input, and the returned geographical communities are more in line with the real scenario.

Description

Technical Field

[0001] The present invention belongs to the field of computer applications, and particularly relates to a geographical community search method based on a time window. Background Art

[0002] In recent years, with the popularization of intelligent mobile devices, users in different geographical locations are getting closer and closer, thus giving rise to various social networks. In a social network, connected users form a community, and their internal connections are close while external connections are sparse (in a computer, a social network is represented by a graph, and a community is represented by a set of edges that meet specific conditions). Analyzing user communities is an important part of social network analysis and research. Since each user contains valuable geographical location information, existing research methods can construct a social network with close connections that meet certain specific conditions geographically for users in different geographical locations, and then, based on the aggregation attributes of the social network, recommend points of interest, form travel teams, etc. for user groups with similar geographical locations. Therefore, the research in this field is of great significance. However, these methods often ignore the timeliness of the connections between users. In real life, users in different geographical locations generally only have close connections within a certain period of time. Summary of the Invention

[0003] The purpose of the present invention is to provide a geographical community search method based on a time window to solve the above technical problems.

[0004] To solve the above technical problems, the specific technical solution of a geographical community search method based on a time window of the present invention is as follows:

[0005] A geographical community retrieval method based on a time window includes the following steps:

[0006] Step 1: Preprocess the geographical social network data set with time attributes, construct the data set into an undirected graph, and simplify the situation where there are multiple edges between the same pair of nodes due to different time attributes. Merge the multiple edges between two nodes into one edge, that is, perform a union operation on the time attributes of each edge to establish a basis for subsequent geographical community search based on a time window;

[0007] Step 2: On the basis of Step 1, obtain all communities in different time periods, and then perform k-core decomposition on the communities belonging to different time periods;

[0008] Step 3: On the basis of Step 1 and Step 2, establish a TDC-index index table structure, and establish a geographical community index table structure based on a time window according to the connections of users with their own close communities in different time periods;

[0009] Step 4: Input the community compactness k, time window [Qts, Qte], and distance threshold s, where Qts and Qte represent the input query start time and end time respectively. Query in the index structure established in Step 3, return the users that meet the time window and compactness, then filter and screen them based on geographical location, iteratively remove the users that do not meet the conditions, and finally return the geographical community that meets the constraint conditions.

[0010] Step 5: Return the query result.

[0011] Furthermore, Step 1 includes the following specific steps:

[0012] For each piece of data (u, v, ti, s) in the dataset, where each piece of data represents an edge, that is, obtain the other users that each user is connected to at all times. Integrate the data of the same pair of connected users at different times into one piece of data (u, v, (ti, tj,...), s), that is, integrate multiple edges between two nodes into one edge.

[0013] Furthermore, Step 2 pre-statistically obtains the user communities of all time windows, obtains all geographical communities by the method of exhaustive time windows, and then performs decomposition processing on them. The specific steps are as follows:

[0014] Based on Step 1, the minimum time and maximum time of the whole graph are ts and te respectively. Using the method of nested loops, the outer loop fixes the start time ts and increments it by 1 each time until ts = tmax, where tmax refers to the value of te after each loop change. For the inner loop of te, continuously decrement it until te = ts. In the loop body, only retain the edges within the time window [ts, te], and perform k-core decomposition on all eligible edges.

[0015] Furthermore, for the given graph G in Step 2, define k-core to represent the cohesion degree, where the value of k is represented as core-number, and the larger the core-number, the higher the cohesion degree. The specific steps are as follows:

[0016] The degree of a node is represented by d. Sequentially select the node v with the smallest degree, assign the degree of all nodes to the current minimum degree d, delete all nodes with degree d and the corresponding edges of the nodes, and at the same time judge whether the degrees of the remaining nodes are greater than d. If not, continue to delete other nodes with degrees less than or equal to d and their adjacent edges until the degrees of all nodes are greater than d. At this time, assign the core-number of the remaining nodes to the minimum degree d' of the remaining nodes, and continuously repeat this process until all nodes are processed.

[0017] Furthermore, in step 3, based on the core-number of each node in all time windows obtained in step 2, and meanwhile, the moment when the core-number of all nodes changes is obtained, which is defined as core-time. For the core-numbers of all nodes in different time periods obtained, the following processing is carried out:

[0018] Sorting of core-number: Starting from the minimum value of core-number, it is judged whether the core-numbers of all nodes v1, v2,..., vn are equal to the currently preset k value in certain time windows. If so, the time windows of a specific node v are further sorted; if not, no processing is performed.

[0019] Sorting of time windows: For all nodes with core-number k, all satisfied time windows are arranged in the form of [ts, te] with ts and te increasing respectively, and all time windows are listed.

[0020] Further process the sorted time windows. Based on the core-time obtained in step 2, that is, when a certain end time te appears, the core-number of the node just meets the current k value, and does not meet at other times te. All time windows that do not meet the conditions are removed. Only when the input query window satisfies Qte >= te and Qts < ts', the core-number of the current node meets the conditions, where te is the termination time of a certain time window, and ts' is the start time ts of the next time window of the time window where te is located.

[0021] At the same time, considering geographical location factors, in the established time index structure, each time window stores the k-th largest distance L (i.e., the distance of the k-th edge) after sorting the geographical distances from the current node to its neighbor nodes in ascending order. Thus, the establishment of TDC-index is completed, laying a foundation for accelerating geographical community retrieval in the future.

[0022] Furthermore, the virtual termination index time temax = te + 1 in step 3 is set to meet the condition that core-number = k in the current time window.

[0023] Furthermore, step 4 includes the following specific steps:

[0024] Step 4-1: Based on step 3, each user node contains time windows that meet the conditions under different core-numbers. By directly inputting the community compactness k and the time window [Qts, Qte], it is judged whether the node meets the compactness and time interval conditions. If it meets the conditions, it is retained; if not, it is directly deleted.

[0025] Step 4-2: For the remaining nodes after the processing in Step 4-1, first find all the nodes whose length L of the k-th edge within the time window is greater than the given distance threshold s. If there are no such nodes, directly return the geographical community composed of all the nodes; if there are such nodes, delete the nodes that do not meet the conditions, and then determine whether the length L of the k-th edge of the neighbor nodes of this node is less than or equal to the distance threshold s. If it is less than or equal to the distance threshold s, keep it; otherwise, delete it. If deleting a certain node does not conform to the cohesive structure of the k-core, directly return non-existence. By this method, iteratively delete the nodes that do not meet the conditions, and finally the community composed of the remaining nodes is the geographical community that meets the conditions to be searched for.

[0026] Further, Step 5 returns the query result, that is, returns the geographical community that meets the conditions based on the time window.

[0027] A method for searching geographical communities based on a time window of the present invention has the following advantages: Through the above method for searching geographical communities based on a time window, the present invention fully explores the correlation between the time window and the user's geographical location, making the queried geographical community more time-sensitive, and pre-establishes an efficient index structure for users within a certain time window and geographical range, providing an efficient query of geographical communities when inputting different time windows, tightness levels, and distance thresholds. Since this invention takes into account the time factor and uses the TDC-index index structure, it has a higher query efficiency and the returned geographical community is more in line with the real scenario. Description of the Drawings

[0028] Figure 1 is the overall flowchart of the present invention.

[0029] Figure 2 is an example of a geographical community with time attributes.

[0030] Figure 3 is an example of a geographical community within a certain time window after processing.

[0031] Figure 4 is the TDC-index index structure.

[0032] Figure 5 is an example of performing geographical threshold processing on users who meet the time window and tightness.

[0033] Figure 6 is the flowchart of the present invention for searching geographical communities based on a time window according to the input query parameters. Detailed Embodiments

[0034] To better understand the purpose, structure, and function of the present invention, the following further describes in detail a method for geographical community search based on a time window of the present invention in conjunction with the accompanying drawings.

[0035] The purpose of the present invention is to overcome the deficiencies of existing research, fully consider the timeliness and spatial distance of the connections between users, and focus on geographical community search based on a time window. Given a social network G=(E, V, T, S), V represents the set of user nodes, E represents the connections between users, that is, the set of edges, T represents the set of moments when there are connections between users in E (i.e., the time attribute of the edges, which is represented by 1, 2, 3... months in the present invention), and S represents the set of distances between the users with connections at a certain moment (i.e., the distance of the edges). For example, the quadruple (ai, aj, ti, s) represents that users ai and aj at different geographical locations have a connection at time ti and their distance is s (s S). The cohesion criterion of the user community used in the present invention is k-core, and an optimized index table structure TDC-index is invented in combination with k-core. Among them, for a community to meet k-core, the following three conditions must be met: (1) The degree of all user nodes in the community is greater than or equal to k (k is an integer), that is, the number of neighbor nodes of each node is greater than or equal to k; (2) The formed community is connected, that is, there do not exist two nodes between which there is no path; (3) The k-core community is a maximally connected graph, that is, there does not exist a community with a number of nodes greater than the number of nodes in the current graph and meeting conditions (1) and (2). Perform k-core decomposition on the communities formed in all different time periods (for example, select the time as January - March, that is, return the communities with edge connections in January, February, and March), and impose geographical distance restrictions (usually, users closer to each other geographically are more willing to participate in offline gatherings and other activities), and an index table structure can be constructed under different k values. The index table can be used to directly retrieve the communities that meet the conditions. The present invention aims to quickly obtain a user geographical community with high cohesion and close geographical location within a certain time period by using the pre-constructed index table structure TDC-index.

[0036] As Figure 1 shown, the present invention proposes a method for geographical community retrieval based on a time window. This method introduces the time attribute of the connections between users and proposes a new index structure to accelerate the query process. The specific technical solution steps are as follows:

[0037] Step 1. Preprocess the geographical social network dataset with time attributes, and construct the dataset into an undirected graph. Due to different time attributes, there may be multiple edges between the same pair of nodes in the graph. In the present invention, a simplification process is carried out, merging multiple edges between two nodes into one edge, that is, performing a union operation on the time attributes of each edge, laying a foundation for subsequent geographical community search based on time windows.

[0038] As Figure 2 shown, for each piece of data (u, v, ti, s) in the dataset, where each piece of data represents an edge in the graph, that is, obtaining other users with whom each user has contact at all times, and integrating the data of the same pair of users with contact at different times into one piece of data (u, v, (ti, tj,...), s), that is, integrating multiple edges between two nodes in the graph into one edge. Figure 2 The figure shows an example of a geographical community with a time window. Among them, node ai represents users in different geographical locations, and the numbers on the edges respectively represent the time t when there is contact between the nodes and their distance s. For example, if user a1 and user a2 have contact in February and their distance is 5 km, then an edge can be constructed between a1 and a2, and the edge has the attribute (2, 5). For example, the distance between user a1 and user a2 at time 1 is 10, denoted as (a1, a2, 1, 10); the distance between user a1 and user a2 at time 2 is also 10, denoted as (a1, a2, 2, 10). Then these two edges can be integrated into one edge, denoted as (a1, a2, (1, 2), 10).

[0039] Figure 3 The figure shows the geographical community graph obtained after Step 1, taking the time window [2, 7] as an example (the time window is constructed for establishing the index table structure, and the user can customize it during query).

[0040] Step 2. Based on Step 1, all communities in different time periods can be obtained (for example, if the minimum time is January and the maximum time is March, communities in the time periods of 1, 2, 3, 1 - 2, 2 - 3, 1 - 3 can be obtained), and then perform k-core decomposition on the communities belonging to different time periods.

[0041] Since there are different connections among community users in different time windows, in order to quickly retrieve the geographical communities that meet the input time window and cohesion conditions, the present invention pre-statistics the user communities of all time windows, obtains all geographical communities by the method of exhausting time windows, and then performs decomposition processing on them. The specific steps are as follows. Based on Step 1, the minimum time (start time) and maximum time (end time) of the whole graph can be obtained as ts and te respectively. Using the method of nested loops, the outer loop fixes the start time ts and increments it by 1 each time until ts = tmax (tmax refers to the value of te after each loop change). For the inner loop of te, it is continuously decreased until te = ts. In the loop body, only the edges within the time window [ts, te] are retained, and k-core decomposition is performed on all qualified edges. In the present invention, for a given graph G, k-core is defined to represent the degree of cohesion, where the value of k is represented as core-number, and the larger the core-number, the higher the degree of cohesion. The specific steps are as follows: The degree of a node is represented by d. The node v with the smallest degree is selected in turn, and the degree of all nodes is given the current minimum degree d. All nodes with degree d and the corresponding edges are deleted. At the same time, it is judged whether the degrees of the remaining nodes are greater than d. If not, other nodes with degrees less than or equal to d and their adjacent edges are continuously deleted until the degrees of all nodes are greater than d. At this time, the core-number of the remaining nodes is assigned the minimum degree d' of the remaining nodes. This process is continuously repeated until all nodes are processed.

[0042] Step 3. On the basis of Step 1 and Step 2, establish a (TimeDistance Core)-index, that is, the TDC-index table structure. According to the connections of users with their own close communities in different time periods, establish a geographical community index table structure based on time windows, which can accelerate the retrieval of geographical communities.

[0043] Based on Step 2, the core-number of each node in all time windows can be obtained, and at the same time, the moment when the core-number of all nodes changes can be obtained, which is defined as core-time. The core-numbers of all nodes in different time periods obtained are processed as follows:

[0044] Sorting of core-number: Starting from the minimum value of core-number, judge whether the core-numbers of all nodes v1, v2,..., vn are equal to the currently preset k value in some time windows. If so, further sort the time windows of a specific node v. If not, no processing is performed.

[0045] Sorting of time windows: For all nodes with core-number k, list all time windows in the form of [ts, te] where ts and te are incremented respectively and all satisfied time windows are arranged in ascending order.

[0046] To reduce the space complexity, further process the sorted time windows. Based on the core-time obtained in step 2, that is, when a certain end time te appears, the core-number of the node just meets the current k value, and does not meet at other times te. Remove all time windows that do not meet the conditions. Taking node a1 as an example, assuming k = 2, after enumerating all time windows, the satisfied time windows {[1, 3], [2, 3], [3, 6], [4, 9]} are obtained. After fixing te, there are many eligible start times ts, and only the smallest eligible start time needs to be stored. When te = 2, there are time windows {[1, 3], [2, 3]} that meet the conditions, and the present invention only needs to store [1, 3]. Only when the input query window satisfies Qte >= te and Qts < ts', the core-number of the current node meets the conditions, where te is the termination time of a certain time window, and ts' is the start time ts of the next time window of the time window where te is located. Since there may be a situation where when te is the maximum value of the time window where it is located, no k-core community that meets the conditions can be found. To ensure the integrity of the index structure, the present invention virtualizes the termination index time temax = te + 1 to make it meet the condition that the core-number = k within the current time window.

[0047] At the same time, considering the geographical location factor, in the established time index structure, each time window stores the k-th largest distance L (i.e., the distance of the k-th edge) after sorting the geographical distances from the current node to its neighbor nodes in ascending order. Thus, the establishment of (Time Distance Core)-index, that is, TDC-index, is completed, laying a foundation for accelerating the retrieval of geographical communities in the future.

[0048] Figure 4 What is shown is based on Figure 2Table of the TDC-index index structure constructed by (a1, a2, a3, a4, a5, a6, a7). Taking node a1 as an example, when k = 2, three time periods are generated, namely [1, 3], [3, 6], and [4, 9]. The end time of each time period is the time when node a1 just meets the core-number, that is, the core-time. The start time of each time period is the minimum time to meet the core-number of the current node. In addition, it is necessary to judge whether the input window time [Qts, Qte] satisfies that Qte is not less than the end time te of a certain time window [ts, te], and then whether Qts is less than the start time ts1 of the adjacent time window [ts1, te1] after the time window [ts, te].

[0049] Assume the input time window is [2, 6] and k = 2. By judging through the index table, 6 is greater than or equal to the end time of the time window [3, 6], and 2 is less than the start time of the subsequent time window [4, 9]. Therefore, v1 meets the conditions of the time window and tightness and is directly returned.

[0050] Step 4: Input the community tightness k, the time window [Qts, Qte] (Qts and Qte represent the input query start time and end time respectively), the distance threshold s, query in the index structure established in step 3, return the users who meet the time window and tightness, then filter and screen them according to the geographical location, iteratively remove the users who do not meet the conditions, and finally return the geographical community that meets the constraint conditions.

[0051] The specific steps are as follows:

[0052] 4-1 Based on step 3, the present invention constructs and obtains the TDC-index. Each user node contains the time windows that meet the conditions under different core-numbers. By inputting the community tightness k and the time window [Qts, Qte], it can be directly judged whether the node meets the tightness and time interval conditions. If it meets the conditions, it is retained; if it does not meet the conditions, it is directly deleted.

[0053] 4-2 For the remaining nodes after being processed in step 4-1, first find all the nodes whose length L of the k-th edge within the time window is greater than the given distance threshold s. If there are none, directly return the geographical community composed of all nodes; if there are, delete the nodes that do not meet the conditions, and judge whether the length L of the k-th edge of the neighbor nodes of the node is less than or equal to the distance threshold s. If it is less than or equal to the distance threshold s, it is retained; otherwise, it is deleted. If deleting a certain node does not conform to the cohesive structure of k-core, directly return that it does not exist. By this method, iteratively delete the nodes that do not meet the conditions, and the community composed of the finally remaining nodes is the geographical community that meets the conditions to be searched.

[0054] Figure 5 The following shows Figure 2 as an example, with the input time window being [2, 7], k = 3 and the threshold being 7, an example of geographical threshold processing through Step 4.

[0055] Figure 6 The following shows the query processes of Steps 3 and 4. Through the input time window [Qts, Qte], the compactness k, and the geographical threshold s, perform k-core decomposition, construct an index table, return user nodes that meet the time window and compactness through the query table, and then iteratively determine whether the returned nodes meet the geographical threshold, and finally output geographical communities that meet the constraints.

[0056] Step 5, return the query result. Returning the query result means returning the geographical community based on the time window that meets the conditions.

[0057] It can be understood that the present invention is described through some embodiments. Those skilled in the art know that without departing from the spirit and scope of the present invention, various changes or equivalent replacements can be made to these features and embodiments. Additionally, under the teaching of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application belong to the scope protected by the present invention.

Claims

1. A geographical community retrieval method based on a time window, characterized in that The steps are as follows: Step 1: Preprocess the geosocial network dataset with time attributes, construct the dataset into an undirected graph, and simplify the situation where there are multiple edges between the same pair of nodes due to different time attributes. Merge the multiple edges between two nodes into one edge, that is, perform a union operation on the time attributes of each edge, laying a foundation for subsequent geographical community search based on time windows. Step 2: On the basis of Step 1, obtain all communities in different time periods, and then perform k-core decomposition on the communities belonging to different time periods. Step 3: On the basis of Step 1 and Step 2, establish a TDC-index index table structure, and establish a geographical community index table structure based on time windows according to the connections of users with their own close communities in different time periods. Step 4: Input the community compactness k, time window [Qts, Qte], and distance threshold s. Qts and Qte represent the input query start time and end time respectively. Query in the index structure established in Step 3, return the users who meet the time window and compactness, and then filter and screen them according to their geographical locations. Iteratively remove the users who do not meet the conditions, and finally return the geographical communities that meet the constraint conditions. Step 5: Return the query result.

2. The method for retrieving a geographical community based on a time window according to claim 1, wherein The specific steps of Step 1 are as follows: For each piece of data (u, v, ti, s) in the dataset, where each piece of data represents an edge, that is, obtain other users with whom each user has connections at all times. Integrate the data of the same pair of connected users at different times into one piece of data (u, v, (ti, tj,...), s), that is, integrate multiple edges between two nodes into one edge.

3. The method for retrieving a geographical community based on a time window according to claim 1, wherein The specific steps of Step 2 are as follows: First, statistically obtain all user communities in time windows, obtain all geographical communities by the method of exhaustive time windows, and then perform decomposition processing on them. Based on Step 1, the minimum time and maximum time of the entire graph are ts and te respectively. Using the method of nested loops, the outer loop fixes the start time ts, and increments it by 1 each time until ts = tmax, where tmax refers to the value of te after each loop change. For the inner loop te, continuously decrease it until te = ts. In the loop body, only retain the edges within the time window [ts, te], and perform k-core decomposition on all eligible edges.

4. The method for retrieving geographical communities based on a time window according to claim 3, wherein For the given graph G in Step 2, define k-core to represent the degree of cohesion, where the value of k is represented as core-number. The larger the core-number, the higher the degree of cohesion. The specific steps are as follows: The degree of a node is represented by d. Sequentially select the node v with the smallest degree, assign the degree of all nodes to the current minimum degree d, delete all nodes with degree d and the corresponding edges. At the same time, judge whether the degrees of the remaining nodes are greater than d. If not, continue to delete other nodes with degrees less than or equal to d and their adjacent edges until the degrees of all nodes are greater than d. At this time, assign the core-number of the remaining nodes to the minimum degree d’ of the remaining nodes. Continuously repeat this process until all nodes are processed.

5. The method for retrieving a geographical community based on a time window according to claim 1, characterized in that, Step 3 obtains the core-number of each node in all time windows based on Step 2, and at the same time obtains the moments when the core-numbers of all nodes change, defined as core-time. For the core-numbers of all nodes in different time periods obtained, the following processing is carried out: Sorting of core-number: Starting from the minimum value of the core-number, judge whether the core-numbers of all nodes v1, v2,..., vn are equal to the currently preset k value in certain time windows. If so, further sort the time windows of a specific node v; if not, do not process. Sorting of time windows: For all nodes with core-number k, arrange all satisfied time windows in the form of [ts, te] and sort them in ascending order of ts and te respectively, and list all time windows. Further process the sorted time windows. Based on the core-time obtained in Step 2, that is, when a certain end time te appears, the core-number of the node just meets the current k value, and it does not meet at other times te. Remove all time windows that do not meet the conditions. Only when the input query window satisfies Qte >= te and Qts < ts’, the core-number of the current node meets the conditions, where te is the termination time of a certain time window, and ts’ is the start time ts of the next time window of the time window where te is located. At the same time, considering the geographical location factor, in the established time index structure, each time window stores the k-th largest distance L (i.e., the distance of the k-th edge) after sorting the geographical distances from the current node to its neighbor nodes in ascending order. Thus, the establishment of the TDC-index is completed, laying a foundation for accelerating the geographical community retrieval in the future.

6. The method for retrieving a geographical community based on a time window according to claim 5, wherein In Step 3, the virtual termination index time temax = te + 1, so that it satisfies the condition that core-number = k in the current time window.

7. The method for retrieving a geographical community based on a time window according to claim 1, characterized in that Step 4 includes the following specific steps: Step 4-1: Based on Step 3, each user node contains time windows that meet the conditions under different core-numbers. Directly judge whether the node meets the compactness and time interval conditions through the input community compactness k and the time window [Qts, Qte]. If it meets the conditions, keep it; if not, directly delete it. Step 4-2: For the remaining nodes after the processing in Step 4-1, first find all the nodes whose length L of the k-th edge within the time window is greater than the given distance threshold s. If there are no such nodes, directly return the geographical community composed of all nodes; If there are such nodes, delete the nodes that do not meet the conditions, and determine whether the length L of the k-th edge of the neighbor nodes of this node is less than or equal to the distance threshold s. If it is less than or equal to the distance threshold s, retain it; otherwise, delete it. If deleting a certain node does not conform to the cohesive structure of the k-core, directly return non-existence. By this method, iteratively delete the nodes that do not meet the conditions. Finally, the community composed of the remaining nodes is the geographical community that meets the conditions to be searched.

8. The method for retrieving a geographical community based on a time window according to claim 1, wherein Step 5 returns the query result, that is, returns the geographical community that meets the conditions based on the time window.

Images