Graph-based frequent ring co-location pattern mining method

Abstract

According to the method, a graph method is introduced, that is, the frequent ring pattern is mined in the graph through graph construction, so that the frequent ring pattern in the data set can be mined more conveniently and quickly. In consideration of space and time attributes of trajectory data, mining work is relatively tedious and complex. Therefore, we choose to convert the trajectory data into transactional data so as to more effectively perform mining. In order to avoid the situation that a large number of redundant modes occur in the mining process, two algorithms are provided to achieve construction of a graph structure and mining of a graph ring. The strategies can effectively filter out the modes which cannot generate the ring co-location mode. Due to the fact that the loop mode has the characteristic of circulation, in the final result set, screening operation is conducted on the mode so as to ensure the accuracy of the result. The mined ring co-location mode can be used in the aspects of tourism recommendation, biologic chains, atmospheric circulation and the like. In addition, in order to verify the accuracy of the mining result, a first synthetic data set specially used for ring co-location mode mining is created.

Description

Technical Field

[0001] The present invention relates to the technical fields of spatial data mining, co-location pattern mining and trajectory pattern mining, and in particular to a graph-based spatial frequent ring pattern mining method. Background Art

[0002] Co-pattern mining is an important branch of spatial data mining. A common co-pattern represents a subset of spatial features whose instances often co-occur in spatial neighborhoods. The goal of spatial co-pattern mining is to find interesting and frequent co-pattern features in various fields. For example, co-patterns can help botanists discover the symbiotic relationship of vegetation in a region, or help medical staff find the source of infection and control the spread of the virus.

[0003] The traditional co-location pattern judgment conditions based on group relations are too strict, resulting in many interesting patterns being ignored. Therefore, we proposed co-location pattern mining based on ring relations. However, because the practical significance of ring co-location patterns based on point data is poor, this paper introduces trajectory data to mine ring patterns.

[0004] Trajectory data is a dataset that records the paths of individuals or objects moving in space. Its value is increasingly prominent as it provides key insights into mobility patterns, route choices, and location preferences. Trajectory mining tasks are mainly divided into trajectory uncertainty, trajectory pattern mining, trajectory classification, and trajectory anomaly detection. In order to reduce the numerous and difficult-to-interpret patterns extracted from sequential pattern algorithms, trajectory patterns impose restrictions on the transition time interval between two consecutive elements and ensure the continuity of patterns along the trajectory. However, this practice inadvertently leads to the neglect of the associations between elements. In contrast, our proposed circular co-location pattern emphasizes the associations between elements while maintaining continuity. It not only helps to avoid the emergence of a large number of complex and difficult-to-interpret patterns in sequential pattern mining, but also helps to alleviate the strict requirements required for trajectory pattern mining.

[0005] Among the trajectory pattern mining algorithms, most of them are expanded on the basis of Apriori-based algorithms. This algorithm uses graph theory knowledge to merge second-order patterns into graph structures and mine ring algorithms in graph structures. The mined ring co-location patterns can be widely used in tourism recommendations, atmospheric circulation, and food chains. Summary of the invention

[0006] The purpose of the present invention is to provide a graph-based spatial frequent ring pattern mining method, aiming to change the current problems of the strictness of satisfying group relationships between instances and the inability to effectively represent path order in point data sets. At the same time, it also constructs the first synthetic dataset specifically for ring co-location pattern mining to solve the problem of lack of relevant ring pattern datasets.

[0007] To achieve the above object, the present invention provides a graph-based frequent ring co-location pattern mining method, comprising the following steps:

[0008] Get the trajectory data of moving objects in a certain period of time;

[0009] Preprocess the trajectory data and transform it into transactional data;

[0010] Traverse the trajectory data set and filter out all frequent second-order linear co-location patterns;

[0011] Through the second-order mode, the head and tail are connected and the graph structure is formed;

[0012] Use the Mining Circular algorithm to mine the rings in the graph structure;

[0013] Determine the frequency of the ring homology pattern;

[0014] Remove redundant frequent ring parity patterns;

[0015] In the process of obtaining the dataset of trajectory data of mobile objects, we first crawl memes from major mainstream media platforms, including Weibo, Qunar, etc., and obtain complete trajectory data in a certain period of time through the GPS trajectory records of mobile objects.

[0016] The data annotation process includes the following steps:

[0017] Normalize the data to remove duplicate data and missing data;

[0018] Then, the trajectory data of the moving objects is cleaned and processed to ensure the accuracy and consistency of the data, and the data is converted into transactional data. The moving paths of each moving object are sorted in chronological order, and the locations are represented by letters to facilitate subsequent mining operations.

[0019] Traverse the moving paths of all moving objects, filter the frequencies of continuous second-order linear co-location patterns, and retain the second-order linear co-location patterns that are greater than or equal to the minimum threshold defined by the user;

[0020] A graph structure can be obtained by first merging frequent second-order patterns. According to anti-monotonicity, if a ring is frequent, then all its sub-patterns must be frequent, so all frequent ring patterns must be in the graph structure;

[0021] The circular pattern in the graph is mined through the Mining circular algorithm. The second-order instances of the mined circular pattern are spliced ​​to obtain the instance of the circular pattern;

[0022] In order to fully mine frequent ring co-location patterns, we compare the frequency of the mined co-location patterns with the minimum threshold defined by the user, and retain the ring patterns that are greater than or equal to the minimum threshold;

[0023] Since the ring pattern itself has the characteristics of a cycle, we further filter the patterns in the result set to remove the redundancy of the results;

[0024] Among them, in order to fully ensure that the pattern in the result set is correct and without redundancy, we finally filter the result set. For example, {A, B, C}, {B, C, A}, {C, A, B} actually essentially express the same ring pattern, but the order of the features is different.

[0025] The present invention provides a graph-based frequent ring co-location pattern mining method, which crawls tourists' travel records from the Internet and organizes them into trajectory data that can be recognized by the program through data cleaning and other methods. At the same time, the trajectory data of each user is sorted according to the time sequence of the places they have visited, and then the trajectory data set is traversed to obtain all frequent second-order patterns, and the Graph-mining algorithm is used for merging operations. According to the merged linear patterns, the ring patterns that meet the conditions are screened out through the verification pattern. Finally, for the completeness and simplicity of the results, we filter the ring patterns once. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Figure 1 It is a flow chart of a graph-based frequent ring co-location pattern mining method of the present invention.

[0028] Figure 2 This is a pseudo code introduction to the Graph-mining algorithm of the present invention.

[0029] Figure 3 This is the pseudo code introduction of the Mining circular algorithm of the present invention.

[0030] Figure 4 It is a pattern instance synthesis graph of the graph-based frequent ring co-location pattern mining method of the present invention. DETAILED DESCRIPTION

[0031] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0032] See also Figure 1 The present invention proposes a graph-based frequent ring co-location pattern mining method, comprising the following steps:

[0033] S1: Obtain the trajectory data of moving objects in a certain period of time;

[0034] S2: Preprocess the trajectory data and convert it into transactional data;

[0035] S3: traverse the trajectory dataset and filter out all frequent second-order linear co-location patterns;

[0036] S4: The second-order mode is connected head to tail and forms a graph structure;

[0037] S5: Use the Mining Circular algorithm to mine the rings in the graph structure;

[0038] S6: Determine the frequency of the ring homology pattern;

[0039] S7: delete redundant frequent ring parity patterns;

[0040] In the process of obtaining the dataset of trajectory data of mobile objects, we first crawl memes from major mainstream media platforms, including Weibo, Qunar, etc., and obtain complete trajectory data in a certain period of time through the GPS trajectory records of mobile objects.

[0041] Clean the crawled data and delete missing data and abnormal data.

[0042] The input data required by this algorithm are trajectory dataset, location features in the dataset and minimum threshold defined by the user.

[0043] The moving paths of all moving objects are traversed, and the frequencies of continuous second-order linear co-location patterns are screened, and the second-order linear co-location patterns that are greater than or equal to the minimum threshold defined by the user are retained.

[0044] By merging frequent second-order linear patterns, linear pattern pruning operations are performed on the generated high-order linear patterns. According to the characteristics of anti-monotonicity, if a linear pattern is not a frequent linear pattern, then the ring pattern it constitutes must not be frequent.

[0045] Not only do we merge the patterns, but we also merge the row instances under the patterns to obtain the row instances of each pattern. At the same time, we introduce a verification pattern to determine whether a linear pattern can form a ring pattern. The verification pattern is composed of the last feature and the first feature of the frequent linear pattern.

[0046] In order to fully mine frequent ring co-location patterns, we compare the frequency of the mined co-location patterns with the user-defined minimum threshold and retain the ring patterns that are greater than or equal to the minimum threshold.

[0047] Since the ring pattern itself has a cyclical characteristic, we further filter the patterns in the result set to remove the redundancy of the results.

[0048] Furthermore, the present invention is further described in detail with reference to specific embodiments and pseudo codes:

[0049] See also Figure 2 ,The specific steps of the algorithm are as follows:

[0050] Algorithm 1 shows the pseudo code for synthesizing the second-order pattern into a graph structure. PCCP represents the frequent ring co-location pattern, RI2 represents the second-order pattern, G represents the generated graph structure, and SCCs represents the connected graph.

[0051] (Step 1) Algorithm 1 shows the pseudo code for mining ABCA of all PCCPs. First, the second-order patterns and their row instances are obtained by traversing the given trajectory dataset.

[0052] (Step 2) Then, all second-order PCCPs are generated according to min_sprev.

[0053] (Step 3) Generate graph G by connecting all second-order patterns end to end.

[0054] (Step 4) Use Tarjan algorithm to divide the graph G into several strongly connected graphs, because a cycle can only appear in a strongly connected graph.

[0055] (Step 5) Traverse each strongly connected graph to obtain the loops in each strongly connected graph.

[0056] (Step 6) Sort the feature points in the strongly connected graph by the number of neighbors, because the more neighbors a point has, the more likely it is to form a ring pattern.

[0057] (Steps 7-10) Traverse each feature and use Algorithm 2 to mine the loops in the strongly connected graph.

[0058] And Algorithm 2( Figure 3 ) is the algorithm 1( Figure 2 ) in detail:

[0059] (Step 1-3) The algorithm mainly finds the loop in the strongly connected graph, pushes all feature points into the stack, and marks the current point as 0.

[0060] (Step 4-11) Traverse the neighbor nodes of the current point. If the neighbor node is equal to the start node, it indicates that a ring pattern has been found. Otherwise, if the neighbor node has been marked, recursive algorithm 2 is used for further mining.

[0061] (Steps 12-15) If find_circular is true, we release the block on the current node, and we release the blocks on all other vertices associated with the current node.

[0062] (Steps 16-18) If a candidate ring pattern is not found, we add all neighbors of the current node as the key of the block and use the neighboring nodes as its associated value.

[0063] (Steps 19-23) Finally, we delete the current node and return true if a candidate cyclic homograph is found.

[0064] Graph-based algorithm graph example Figure 4 As shown in the figure, the specific steps of generating the ring pattern {A, B, E} are explained in combination with the schematic diagram:

[0065] Step 1: Split the {A,B,E} ring pattern into {A,B}, {B,E}, {E,A}.

[0066] Step 2: Get instances of the {A,B}, {B,E}, and {E,A} patterns, and get the ABE ring instance by merging the head and tail of the instances.

[0067] Step 3: Compare the frequency of the pattern with the minimum threshold. If it is greater than the minimum threshold, the pattern is considered frequent.

[0068] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A graph-based frequent ring co-location pattern mining method, characterized in that: The following steps are involved: Get the trajectory data of moving objects in a certain period of time; Preprocess the trajectory data and transform it into transactional data; Traverse the trajectory data set and filter out all frequent second-order linear co-location patterns; Through the second-order mode, the head and tail are connected and the graph structure is formed; Use the Mining Circular algorithm to mine the rings in the graph structure; Determine the frequency of ring homology patterns; Remove redundant frequent ring parity patterns.

2. The graph-based frequent ring co-location pattern mining method according to claim 1, characterized in that: In the process of obtaining the dataset of trajectory data of mobile objects, we first crawl memes from major mainstream media platforms, including Weibo, Qunar, etc., and obtain complete trajectory data in a certain period of time through the GPS trajectory records of mobile objects.

3. The graph-based frequent ring co-location pattern mining method according to claim 1, characterized in that: The trajectory data of mobile objects is cleaned and processed to ensure the accuracy and consistency of the data, and the data is converted into transactional data. The movement path of each mobile object is sorted in chronological order, and the locations are represented by letters to facilitate subsequent mining operations.

4. The graph-based frequent ring co-location pattern mining method according to claim 1, characterized in that: The moving paths of all moving objects are traversed, and the frequencies of continuous second-order linear co-location patterns are screened, and the second-order linear co-location patterns that are greater than or equal to the minimum threshold defined by the user are retained.

5. The graph-based frequent ring co-location pattern mining method according to claim 1, characterized in that: A graph structure can be obtained by first merging frequent second-order patterns. According to anti-monotonicity, if a ring is frequent, then all its sub-patterns must be frequent, so all frequent ring patterns must be in the graph structure.

6. The graph-based frequent ring co-location pattern mining method according to claim 1, characterized in that: The mining-circular algorithm is used to mine the circular patterns in the graph. The second-order instances of the mined circular patterns are spliced ​​to obtain the instances of the circular patterns.

7. The graph-based frequent ring co-location pattern mining method according to claim 1, characterized in that: In order to fully mine frequent ring co-location patterns, we compare the frequency of the mined co-location patterns with the user-defined minimum threshold and retain the ring patterns that are greater than or equal to the minimum threshold.

8. The graph-based frequent ring co-location pattern mining method according to claim 1, characterized in that: Since the ring pattern itself has a cyclical characteristic, we further filter the patterns in the result set to remove the redundancy of the results.

Images