Subway and bus combined travel space-time pattern recognition method based on two-step clustering

Abstract

The invention relates to the technical field of spatio-temporal data mining, in particular to a metro and bus combination travel spatio-temporal pattern recognition method based on two-step clustering, which comprises the following steps: defining transfer spatial features according to spatial features of two sides of a metro and bus appearing in combination; based on transfer space characteristics, a comprehensive transfer similarity space measurement method is provided; a two-step clustering method is designed to identify a transfer space-time mode of combined travel; and providing a transfer station function feature extraction method to verify a transfer mode. According to the method, the spatial features of the two sides of a subway and a bus in combined travel are combined, a two-step clustering algorithm based on comprehensive transfer similarity is provided, the spatial features of the combined travel can be better captured, and a transfer space-time mode is effectively recognized.

Description

Technical Field

[0001] The invention relates to the technical field of spatiotemporal data mining, and in particular to a method for identifying spatiotemporal patterns of subway and bus combined travel based on two-step clustering. Background Art

[0002] The convenience of transfer is the key to the attractiveness of the combination of subway and regular bus travel, and its smoothness and efficiency will greatly affect the travel experience of passengers. Therefore, by understanding the spatiotemporal characteristics of the combination of subway and bus travel and identifying the spatiotemporal laws of transfer travel, the organic connection between subway and bus can be strengthened, the service quality of public transportation can be improved, and its competitive advantage compared with private transportation can be enhanced.

[0003] Many studies have classified subway stations according to the similarity of passengers in a specific time period, but considering that transfer travel data also contains the spatial information of subways and buses, even adjacent transfer stations may show complex spatiotemporal heterogeneity in carrying combined travel due to the different distribution of starting points and destinations, making the spatiotemporal pattern recognition results of existing methods less accurate. Summary of the invention

[0004] In order to solve the technical problem that the spatiotemporal pattern recognition results of the existing methods are of low accuracy, the purpose of the present invention is to provide a spatiotemporal pattern recognition method for subway and bus combined travel based on two-step clustering. The technical scheme adopted is as follows:

[0005] In the time interval corresponding to each time interval, the spatial characteristics of each transfer time and space station are determined according to the spatial characteristics of the subway side and the spatial characteristics of the bus side of each transfer station;

[0006] According to the spatial characteristics of different transfer time and space stations, the information similarity under the line distribution of the spatial characteristics of the bus side and the information similarity under the grid distribution of the spatial characteristics of the subway side are obtained to obtain the transfer space similarity between different transfer time and space stations;

[0007] According to the transfer space similarity between different transfer space-time sites corresponding to different time intervals, the transfer space-time sites are clustered in the first step to obtain the first step clustering result, wherein each cluster in the first step clustering result corresponds to a transfer space mode;

[0008] In the first step of clustering results, according to the starting and ending demand of each transfer space-time station, combined with the similarity between the transfer space pattern time series of each transfer station and other transfer stations, the transfer stations are clustered in the second step to obtain the space-time pattern recognition results of the second clustering.

[0009] Preferably, the spatial characteristics of the transfer time and space station are expressed as t∈{1,2,…,T},i∈{1,2,…,I}, where represents the spatial characteristics of the i-th transfer station in the t-th time interval, T represents the total number of time intervals, I represents the total number of transfer time and space stations, represents the feature set of each subway station that transfers at the i-th transfer time and space station in the t-th time interval, Represents the feature set of bus routes that transfer at the i-th transfer time and space station in the t-th time interval.

[0010] Preferably, the transfer space similarity between different transfer time and space stations is obtained based on the information similarity of the spatial characteristics of different transfer time and space stations under the line distribution of the bus side spatial characteristics and the information similarity of the spatial characteristics of the subway side under the grid distribution, specifically including:

[0011] Based on the cosine similarity between the spatial characteristics of each two transfer time and space stations on the subway side, the subway spatial distribution similarity between each two transfer time and space stations is determined;

[0012] Based on the cosine similarity between the spatial characteristics of each two transfer time and space stations on the bus side, the similarity of the bus spatial distribution between each two transfer time and space stations is determined;

[0013] The subway spatial distribution similarity and the bus spatial distribution similarity are weightedly summed to obtain the transfer spatial similarity between every two transfer time-space stations.

[0014] Preferably, the first step of clustering the transfer space-time sites according to the transfer space similarities between different transfer space-time sites corresponding to different time intervals to obtain the first step clustering results specifically includes:

[0015] The negative correlation coefficient of the transfer space similarity between different transfer time and space stations is used as the transfer time and space distance between different transfer time and space stations;

[0016] According to the transfer space similarity between different transfer time and space stations, the local time and space density and the corresponding minimum time and space distance are obtained;

[0017] Based on the transfer time-space distance, the local time-space density and the corresponding minimum time-space distance are taken, and the transfer time-space sites are clustered in the first step using the TCS-CFSFDP algorithm to obtain the first step clustering result.

[0018] Preferably, the obtaining of the local space-time density and the corresponding minimum space-time distance according to the transfer space similarity between different transfer space-time stations specifically includes:

[0019]

[0020] in, represents the local space-time density of the transfer time-space station corresponding to the i-th transfer station in the t-th time interval, j represents the j-th transfer station, m represents the m-th time interval, represents the transfer time-space distance between the i-th transfer station in the t-th time interval and the j-th transfer station in the m-th time interval, and dst represents the spatial cutoff distance; represents the minimum space-time distance between the i-th transfer station and the corresponding transfer time-space station in the t-th time interval, It represents the local space-time density of the transfer space-time station corresponding to the j-th transfer station in the m-th time interval.

[0021] Preferably, in the first step clustering result, according to the starting and ending demand of each transfer time and space station, combined with the similarity between the transfer space pattern time series of each transfer station and other transfer stations, the transfer stations are clustered in the second step, which specifically includes:

[0022] In the first clustering result, the transfer space-time stations corresponding to the number of starting and ending demands greater than the first preset threshold are classified as transfer space patterns with high demand values, and the transfer space-time stations corresponding to the number of starting and ending demands less than the second preset threshold are classified as transfer space patterns with extremely low demand values; further, the time series of each transfer station is obtained, and based on the cosine similarity of the time series between different transfer stations, the AP algorithm is used to perform the second step clustering of the transfer stations.

[0023] Preferably, the method for obtaining the attractiveness and belongingness in the AP algorithm specifically includes:

[0024] r t (i,j)=λr t-1 (i,j)+(1-λ)r t (i,j)

[0025] a t (i,j)=λa t-1 (i,j)+(1-λ)a t (i,j)

[0026] Among them, r t (i,j) and a t (i, j) respectively represent the attraction and belonging between the i-th transfer time and space station and the j-th transfer time and space station in the t-th round, λ is the damping factor, r t (i,j) and a t-1 (i, j) respectively represent the attraction and belonging degree between the i-th transfer time-space station and the j-th transfer time-space station in the t-1th round.

[0027] Preferably, the method further comprises:

[0028] Obtaining the POI distribution vectors of all transfer stations at each transfer station to form a POI distribution matrix for each transfer station; wherein the POI distribution vector of each station is composed of the number of POIs of various types in the buffer zone of each station;

[0029] The ratio between the demand of each station transferred at each transfer station and the starting and ending point demand of the corresponding transfer station is used as the weight of each station transferred at each transfer station;

[0030] The POI distribution matrix is ​​weighted by using the weights to obtain the POI distribution of each transfer station, and the transfer function characteristics of each transfer station are obtained according to the POI distribution.

[0031] Preferably, the acquiring the transfer function characteristics of each transfer station according to the POI distribution specifically includes:

[0032]

[0033] in, represents the transfer function characteristics of the p-th POI type in the i-th transfer station buffer, represents the POI distribution of the pth POI type in the buffer zone of the i-th transfer station, N i represents the total number of transfer POIs at the i-th transfer station, N P Represents the sum of the number of all POI types in the study area, and N is the sum of the number of all POIs in the study area.

[0034] The embodiments of the present invention have at least the following beneficial effects:

[0035] The present invention first defines the transfer space characteristics according to the spatial characteristics on both sides of the combined subway and bus; then, based on the transfer space characteristics, a comprehensive transfer similarity spatial measurement method is proposed; a two-step clustering method is further designed to identify the transfer spatiotemporal pattern of combined travel; in addition, a transfer station functional feature extraction method is proposed to verify the transfer pattern. The present invention combines the spatial characteristics of both sides of the subway and bus in the combined travel, and proposes a two-step clustering algorithm based on comprehensive transfer similarity, which can better capture the spatial characteristics of the combined travel and effectively identify the transfer spatiotemporal pattern. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are 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.

[0037] Figure 1 It is a flowchart of the steps of a method for identifying the spatiotemporal pattern of subway and bus combined travel based on two-step clustering provided by the present invention;

[0038] Figure 2 It is a schematic diagram of OD demand distribution of a transfer station provided by the present invention;

[0039] Figure 3 It is a schematic diagram of the AP clustering model structure provided by the present invention;

[0040] Figure 4 It is a schematic diagram of the proportion of bus segment line demand in transfer space mode 0 provided by the present invention;

[0041] Figure 5 It is a schematic diagram of the proportion of bus segment line demand in the transfer space mode 1 provided by the present invention;

[0042] Figure 6 It is a schematic diagram of the proportion of bus segment line demand in the transfer space mode 2 provided by the present invention;

[0043] Figure 7 It is a schematic diagram of the proportion of bus segment line demand in the transfer space mode 3 provided by the present invention;

[0044] Figure 8 It is a schematic diagram of the proportion of bus segment line demand of transfer space mode 4 provided by the present invention;

[0045] Fig. 9 It is a schematic diagram of the time series changes of the transfer time series mode 0 provided by the present invention. DETAILED DESCRIPTION

[0046] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a subway bus combined travel spatiotemporal pattern recognition method based on two-step clustering proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0047] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0048] The following is a detailed description of a method for identifying spatiotemporal patterns of subway and bus combined travel based on two-step clustering provided by the present invention in conjunction with the accompanying drawings.

[0049] See also Figure 1 , which shows a flowchart of a method for identifying spatiotemporal patterns of subway and bus combined travel based on two-step clustering provided by an embodiment of the present invention, the method comprising the following steps:

[0050] Step 1: In the time interval corresponding to each time interval, the spatial characteristics of each transfer time and space station are determined according to the spatial characteristics of the subway side and the spatial characteristics of the bus side of each transfer station.

[0051] First, based on the spatial characteristics of the combination of subways and buses on both sides, the transfer space characteristics are defined. The basic research unit of the present invention is the transfer station S on the time interval t∈{1,2,…,T} i ∈{S 1 ,S 2 ,…,S I}, the transfer time and space station is expressed as in It represents the transfer time-space station corresponding to the i-th transfer station in the t-th time interval, T represents the total number of time intervals, and I represents the total number of transfer time-space stations.

[0052] In this embodiment, the time length of each time interval is 30 minutes, and the implementer can set it according to the specific implementation scenario. The selection of transfer stations can be based on the demand for transfers, and 200 transfer stations are selected in descending order of demand for subsequent analysis, that is, in this embodiment, the value of I is 200.

[0053] like Figure 2 As shown, the transfer time and space station Spatial characteristics Space characteristics of subway side and bus side space characteristics Composition, expressed as in represents the spatial characteristics of the i-th transfer station in the t-th time interval, represents the feature set of each subway station that transfers at the i-th transfer time and space station in the t-th time interval, Represents the feature set of bus routes that transfer at the i-th transfer time and space station in the t-th time interval.

[0054] Furthermore, the spatial characteristics of the subway side By transfer station The distribution characteristics of the combined demand of each subway station where the transfer takes place can be expressed as e∈(1,2,…,E), where represents the demand for subway station e at the i-th transfer station in the t-th time interval, and E represents the total number of subway stations.

[0055] Bus side space characteristics By transfer station Combined demand for each bus line with transfers The distribution characteristics of l∈(1,2,…,L), where L represents the total number of bus routes, represents the demand for bus line l at the i-th transfer station in the t-th time interval.

[0056] Combined demand based on the number of stops passengers take on a bus route Can be further divided into and Expressed as represents the demand for the first category road of the lth bus line at the i-th transfer station in the t-th time interval. Figures 4 to 8 As shown, a schematic diagram of the proportion of bus segment line demand in the five transfer space modes provided in this embodiment is shown, namely, transfer space mode 0, transfer space mode 1, transfer space mode 2, transfer space mode 3, and transfer space mode 4.

[0057] like Figure 2 As shown in Figure 1, if a passenger takes 3 stops or less on the lth bus line, he is classified as If the number of stops is between 4 and 10 (inclusive), it is classified as If the number of stops is more than 10, it is classified as

[0058] Step 2: According to the spatial characteristics of different transfer time and space stations, the information similarity under the line distribution of the spatial characteristics of the bus side and the information similarity under the grid distribution of the spatial characteristics of the subway side are respectively obtained to obtain the transfer space similarity between different transfer time and space stations.

[0059] First, it should be noted that this embodiment uses transfer similarity to perform spatial measurement based on transfer space characteristics, and in order to simplify the calculation dimension of subway spatial distribution similarity, the subway network is divided into multi-scale spatial grids. Specifically, the subway coverage area of ​​the city is divided into 9 grids with a side length of 10 kilometers in the central urban area and 8 grids with a side length of 30 kilometers in the suburbs.

[0060] Then, the spatial characteristics of the subway side It can be further expressed as the distribution of combined demand within each subway grid, expressed as g∈(1,2,…,G), where G represents the total number of grids, It represents the demand of the i-th transfer station in the g-th subway grid in the t-th time interval. Based on the cosine similarity between the spatial characteristics of each two transfer time-space stations on the subway side, the subway spatial distribution similarity between each two transfer time-space stations is determined.

[0061] As a specific example, the similarity of subway spatial distribution can be expressed as:

[0062]

[0063] in, represents the subway spatial distribution similarity between the corresponding transfer time-space stations of the i-th transfer station in the t-th time interval and the j-th transfer station in the m-th time interval, j represents the j-th transfer station, m represents the m-th time interval, represents the demand for the jth transfer station in the gth subway grid during the mth time interval.

[0064] According to the same acquisition method as the subway spatial distribution similarity, based on the cosine similarity between the spatial characteristics of each two transfer time and space stations on the bus side, the bus spatial distribution similarity between each two transfer time and space stations is determined, which can be expressed by the formula:

[0065]

[0066] in, represents the similarity of the spatial distribution of public transportation between the i-th transfer station in the t-th time interval and the j-th transfer station in the m-th time interval, f represents the category of the bus line, l represents the l-th bus line, represents the demand for the f-th category road of the l-th bus line at the i-th transfer station in the t-th time interval, It represents the demand for the f-th category road of the l-th bus line at the j-th transfer station in the m-th time interval.

[0067] Furthermore, the subway spatial distribution similarity and the bus spatial distribution similarity are weighted and summed to obtain the transfer space similarity between every two transfer time and space stations. The calculation formula can be expressed as:

[0068]

[0069] in, represents the transfer space similarity between the i-th transfer station in the t-th time interval and the j-th transfer station in the m-th time interval, ω 1 and ω 2 They represent the weights of the similarity of subway spatial distribution and bus spatial distribution, respectively.

[0070] Step three, based on the transfer space similarity between different transfer space-time stations corresponding to different time intervals, the transfer space-time stations are clustered in the first step to obtain the first step clustering results, in which each cluster cluster corresponds to a transfer space mode.

[0071] This embodiment is based on a two-step clustering method to identify the transfer space-time pattern of subway and bus combined travel. On the first aspect, the transfer space pattern recognition based on the TCS-CFSFDP algorithm is first introduced.

[0072] In this embodiment, a clustering algorithm framework for fast search and density peak discovery based on transfer comprehensive similarity (TCS-CFSFDP) is used to identify the transfer spatial pattern of each transfer station in different time intervals.

[0073] First, the negative correlation coefficient of the transfer space similarity between different transfer space-time stations is used as the transfer space-time distance between different transfer space-time stations. In this embodiment, the transfer space-time distance can be expressed as: represents the transfer time-space distance between the i-th transfer station in the t-th time interval and the j-th transfer station in the m-th time interval. Represents the corresponding transfer space similarity.

[0074] Then, in the TCS-CFSFDP algorithm, there are two important parameters for any transfer time-space station, namely the local time-space density and the minimum time-space distance between it and the higher density point, and the acquisition method is based on the comprehensive similarity characteristics of the transfer.

[0075] Specifically, the local space-time density and the corresponding minimum space-time distance are obtained according to the transfer space similarity between different transfer space-time stations; based on the transfer space-time distance, the local space-time density and the corresponding minimum space-time distance are taken, and the transfer space-time stations are clustered in the first step using the TCS-CFSFDP algorithm to obtain the first step clustering result.

[0076] Among them, the local space-time density and the corresponding minimum space-time distance can be expressed by the formula:

[0077]

[0078] in, represents the local space-time density of the transfer time-space station corresponding to the i-th transfer station in the t-th time interval, j represents the j-th transfer station, m represents the m-th time interval, represents the transfer time-space distance between the i-th transfer station in the t-th time interval and the j-th transfer station in the m-th time interval, and dst represents the spatial cutoff distance; represents the minimum space-time distance between the i-th transfer station and the corresponding transfer time-space station in the t-th time interval, It represents the local space-time density of the transfer space-time station corresponding to the j-th transfer station in the m-th time interval.

[0079] In this embodiment, the spatial cutoff distance is set to the 3% quartile of the transfer time and space distance. The cluster center is determined based on the point in the decision graph with the largest local time and space density and the smallest time and space distance between it and the higher density point. The cluster center is usually identified as a point with a large and The points with the same value. Each transfer time and space station is matched to its own cluster center to obtain 5 transfer space modes. That is, the first step clustering result in this embodiment includes 5 transfer space modes. By counting the subway demand distribution, bus segment line demand proportion and OD distribution of each transfer space mode, its start and end point (OD) distribution characteristics can be obtained.

[0080] Step 4: Based on the clustering results of the first step, according to the starting and ending demand of each transfer time and space station, combined with the similarity between the transfer space pattern time series of each transfer station and other transfer stations, the transfer stations are clustered in the second step to obtain the spatiotemporal pattern recognition results of the second clustering.

[0081] Based on the transfer space pattern identified in the first step, the transfer time and space stations corresponding to the number of starting and ending demands greater than the first preset threshold are classified as transfer space patterns with high demand values, and the transfer time and space stations corresponding to the number of starting and ending demands less than the second preset threshold are classified as transfer space patterns with extremely low demand values. In this embodiment, the value of the first preset threshold is 100, and the value of the second preset threshold is 10. The implementer can set it according to the specific implementation scenario in other embodiments.

[0082] Specifically, the starting and ending points of the transfer The transfer space pattern of the transfer time and space station is reclassified into the transfer space pattern with high demand value, and the demand for the starting and ending points of the transfer that was previously eliminated is The transfer time and space stations are divided into transfer space modes with extremely low demand values. represents the starting and ending demand of the transfer at the i-th transfer station in the t-th time interval. Finally, 11 transfer space patterns are obtained, and the AP algorithm is used to further identify the transfer time series pattern. The steps are as follows: Figure 3 shown.

[0083] Then, the transfer space mode time series of each transfer station is obtained. In this embodiment, the transfer space mode time series TP of the i-th transfer station is i It can be expressed as in It represents the transfer space pattern of the ith transfer station in the tth time interval, based on the time series composed of the transfer space pattern of the ith transfer station in the time interval {1,2,…,T} in the first step clustering result, specifically the 5 transfer space patterns shown in step 3, and after analysis in combination with the start and end point demands, it includes 11 transfer space patterns, which are composed of the numbers corresponding to the transfer space patterns, for example, the exemplary transfer space pattern 0 in step 3, and the cosine similarity of the transfer space pattern time series of different transfer stations is selected to describe the similarity between them, that is, the cosine similarity of the time series between different transfer stations is obtained.

[0084] Furthermore, the time series of the transfer space pattern of each transfer station is mapped to a data point in the T-dimensional space to transmit two types of information, namely, attraction and belonging. It should be noted that, in this embodiment, the AP (Attraction and Affiliation) algorithm is used for the second step of clustering analysis. The AP algorithm is a clustering method, which identifies the cluster center and performs clustering by calculating attraction and belonging.

[0085] In this embodiment, the attraction r i,j It means that after considering other potential cluster centers, the transfer station S j As a transfer station S i The degree of adaptation of the cluster center. i,j Indicates that other transfer stations are considered for transfer station S j After becoming the support of the cluster center, the transfer station S i Select transfer station S j As the adaptation degree of the cluster center, its initial value is 0.

[0086] Specifically, the attractiveness and belongingness can be expressed using the calculation formula:

[0087]

[0088] Among them, r(i,j) represents the transfer station S j For transfer station S i The attractiveness reflects the updating process of attractiveness and indicates the transfer station S i According to its j The similarity of and the degree of belonging to other stations are adjusted to select the most appropriate cluster center. j′ represents the j′th transfer station, that is, the transfer station S j′ , TS(i,j′) represents the transfer station S j and transfer station S j′ The cosine similarity between .

[0089]

[0090] Where a(i,j) represents the transfer station S j For transfer station S i The degree of belonging reflects the transfer station S i Whether to support transfer station S j is called the degree of adaptation of its cluster center. i′ represents the i′th transfer station, that is, the transfer station S i′ , r(i′,j) represents the transfer station S j For transfer station S i′ attractiveness.

[0091] Repeat the above-mentioned updating process of attraction and belongingness to continuously update the attraction r i,j and degree of belonging i,j To ensure the stability of information update, the damping factor λ is set to partially retain the information of the previous moment. Specifically, the calculation formula of attraction and belonging in the AP algorithm can be expressed as:

[0092] r t (i,j)=λr t-1 (i,j)+(1-λ)r t (i,j)

[0093] a t (i,j)=λa t-1 (i,j)+(1-λ)a t (i,j)

[0094] Among them, r t (i,j) and a t (i, j) respectively represent the attraction and belonging between the i-th transfer time and space station and the j-th transfer time and space station in the t-th round, λ is the damping factor, r t (i,j) and a t-1 (i, j) respectively represent the attraction and belonging degree between the i-th transfer time-space station and the j-th transfer time-space station in the t-1th round.

[0095] When the algorithm converges, the attraction and belonging degrees are added to obtain the cluster center. i,j +r i,j The maximum j value is j', then when i = j', the transfer station S j is the cluster center; when i≠j′, the transfer station S i Belongs to cluster center S j, .

[0096] Then, the time series vectors of each cluster after the cluster center assignment are restored to the transfer space pattern time series of the corresponding transfer station and the spatial pattern distribution is statistically analyzed. In this embodiment, 7 transfer time series patterns are obtained. The temporal change rules of the transfer space patterns contained in the transfer stations with the same transfer time series pattern are summarized and statistically analyzed to obtain the final spatiotemporal pattern recognition results of the combined trip, such as Fig. 9 A schematic diagram of the time series changes of transfer time series mode 0 in this embodiment is shown.

[0097] Finally, this embodiment also provides an exemplary transfer station functionality extraction method to verify the transfer mode, that is, the weighted POI data of all the starting and ending buffers of the transfer station are used to reflect the functional characteristics of the transfer station. The transfer station functional characteristics extraction method specifically includes:

[0098] Step S1, obtaining POI distribution vectors of all transfer stations at each transfer station to form a POI distribution matrix of each transfer station; wherein the POI distribution vector of each station is composed of the number of POIs of various types in the buffer zone of each station.

[0099] Transfer Station S i The POI distribution matrix is ​​composed of i The POI distribution vectors of all transfer stations are composed of: poi i =[(poi i,1 ) T ,(poi i,2 ) T ,…,(poi i,z ) T ,…,(poi i,Z ) T ] P×Z ,in(·) T is the transpose of the vector. i The POI distribution of station z with a transfer is composed of the number of POIs of various types in the buffer zone of station z, as shown in Eq. p∈(1,2,…P). Among them, p represents the pth POI type, P represents the transfer station S i The total number of POI types in the buffer zone of the transfer station z, Indicates that at the transfer station S i The number of POIs of the pth POI type in the buffer zone of the transfer station z.

[0100] Step S2: taking the ratio between the demand of each transfer station at each transfer station and the starting and ending point demands of the corresponding transfer station as the weight of each transfer station at each transfer station.

[0101] The distribution of the starting and ending points of the transfer at the transfer station is uneven, and the functional characteristics of the station with large demand can better reflect the functional characteristics of the transfer station. i The weight vector W i By in S i The weight vector is composed of all the transfer stations. in Indicates the transfer station S i The weight of the transfer station z. The weight is composed of its demand ratio, which can be expressed as: D i Indicates that at the transfer station S i The starting and ending points of the transfer, Indicates that at the transfer station S i The demand for transfer station z.

[0102] Step S3: weighting the POI distribution matrix using the weight to obtain the POI distribution of each transfer station, and obtaining the transfer function characteristics of each transfer station according to the POI distribution.

[0103] Specifically, the transfer POI distribution, i.e., transfer_poi, can be obtained by calculating the POI distribution matrix and the weight vector, which can be specifically expressed as: Where transfer_poi(i) represents the transfer station S i POI distribution, It is the transfer station S i The number of the p-th POI type weighted by demand.

[0104] Finally, by using the enrichment factor of POI to represent the functional characteristics of the transfer station, the transfer station S can be obtained. i The transfer function characteristics can be expressed as follows:

[0105]

[0106] in, represents the transfer function characteristics of the p-th POI type in the i-th transfer station buffer, represents the POI distribution of the pth POI type in the buffer zone of the i-th transfer station, N i represents the total number of transfer POIs at the i-th transfer station, N P Represents the sum of the number of all POI types in the study area, and N is the sum of the number of all POIs in the study area.

[0107] Furthermore, the transfer station S i Transfer function characteristics EF iIt can be expressed as Then, by taking the mean value, the functional characteristic distribution of each transfer time series pattern is obtained.

[0108] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for identifying spatiotemporal patterns of subway and bus combined travel based on two-step clustering, characterized in that: The method comprises the following steps: In the time interval corresponding to each time interval, the spatial characteristics of each transfer time and space station are determined according to the spatial characteristics of the subway side and the spatial characteristics of the bus side of each transfer station; According to the spatial characteristics of different transfer time and space stations, the information similarity under the line distribution of the spatial characteristics of the bus side and the information similarity under the grid distribution of the spatial characteristics of the subway side are obtained to obtain the transfer space similarity between different transfer time and space stations; According to the transfer space similarity between different transfer space-time sites corresponding to different time intervals, the transfer space-time sites are clustered in the first step to obtain the first step clustering result, wherein each cluster in the first step clustering result corresponds to a transfer space mode; In the first step of clustering results, according to the starting and ending demand of each transfer space-time station, combined with the similarity between the transfer space pattern time series of each transfer station and other transfer stations, the transfer stations are clustered in the second step to obtain the space-time pattern recognition results of the second clustering.

2. The method for identifying spatiotemporal patterns of subway-bus combination travel based on two-step clustering according to claim 1 is characterized in that: The spatial characteristics of the transfer time and space station are expressed as t∈{1,2,…,T},i∈{1,2,…,I}, where represents the spatial characteristics of the i-th transfer station in the t-th time interval, T represents the total number of time intervals, I represents the total number of transfer time and space stations, represents the feature set of each subway station that transfers at the i-th transfer time and space station in the t-th time interval, Represents the feature set of bus routes that transfer at the i-th transfer time and space station in the t-th time interval.

3. The method for identifying spatiotemporal patterns of subway-bus combination travel based on two-step clustering according to claim 1 is characterized in that: The transfer space similarity between different transfer time and space stations is obtained based on the information similarity of the line distribution of the spatial characteristics of the bus side and the information similarity of the grid distribution of the spatial characteristics of the subway side, specifically including: Based on the cosine similarity between the spatial characteristics of each two transfer time and space stations on the subway side, the subway spatial distribution similarity between each two transfer time and space stations is determined; Based on the cosine similarity between the spatial characteristics of each two transfer time and space stations on the bus side, the similarity of the bus spatial distribution between each two transfer time and space stations is determined; The subway spatial distribution similarity and the bus spatial distribution similarity are weightedly summed to obtain the transfer spatial similarity between every two transfer time and space stations.

4. The method for identifying spatiotemporal patterns of subway-bus combination travel based on two-step clustering according to claim 1, characterized in that: The first step of clustering the transfer time and space sites according to the transfer space similarities between different transfer time and space sites corresponding to different time intervals to obtain the first step clustering results specifically includes: The negative correlation coefficient of the transfer space similarity between different transfer time and space stations is used as the transfer time and space distance between different transfer time and space stations; According to the transfer space similarity between different transfer space-time stations, the local space-time density and the corresponding minimum space-time distance are obtained; Based on the transfer time-space distance, the local time-space density and the corresponding minimum time-space distance are taken, and the transfer time-space sites are clustered in the first step using the TCS-CFSFDP algorithm to obtain the first step clustering result.

5. The method for identifying spatiotemporal patterns of subway-bus combination travel based on two-step clustering according to claim 4 is characterized in that: The obtaining of the local space-time density and the corresponding minimum space-time distance according to the transfer space similarity between different transfer space-time stations specifically includes: in, represents the local space-time density of the transfer time-space station corresponding to the i-th transfer station in the t-th time interval, j represents the j-th transfer station, m represents the m-th time interval, represents the transfer time-space distance between the i-th transfer station in the t-th time interval and the j-th transfer station in the m-th time interval, and dst represents the spatial cutoff distance; represents the minimum space-time distance between the i-th transfer station and the corresponding transfer time-space station in the t-th time interval, It represents the local space-time density of the transfer space-time station corresponding to the j-th transfer station in the m-th time interval.

6. The method for identifying spatiotemporal patterns of subway-bus combination travel based on two-step clustering according to claim 1, characterized in that: In the first step clustering results, according to the starting and ending demand of each transfer time and space station, combined with the similarity between the transfer space pattern time series of each transfer station and other transfer stations, the transfer stations are clustered in the second step, specifically including: In the first clustering result, the transfer space-time stations corresponding to the number of starting and ending demands greater than the first preset threshold are classified as transfer space patterns with high demand values, and the transfer space-time stations corresponding to the number of starting and ending demands less than the second preset threshold are classified as transfer space patterns with extremely low demand values; further, the time series of each transfer station is obtained, and based on the cosine similarity of the time series between different transfer stations, the AP algorithm is used to perform the second step clustering of the transfer stations.

7. The method for identifying spatiotemporal patterns of subway-bus combination travel based on two-step clustering according to claim 6 is characterized in that: The method for obtaining the attractiveness and belongingness in the AP algorithm specifically includes: r t (i,j)=λr t-1 (i,j)+(1-λ)r t (i,j) a t (i,j)=λa t-1 (i,j)+(1-λ)a t (i,j) Among them, r t (i,j) and a t (i, j) respectively represent the attraction and belonging between the i-th transfer time and space station and the j-th transfer time and space station in the t-th round, λ is the damping factor, r t (i,j) and a t-1 (i, j) respectively represent the attraction and belonging degree between the i-th transfer time-space station and the j-th transfer time-space station in the t-1th round.

8. The method for identifying spatiotemporal patterns of subway-bus combination travel based on two-step clustering according to claim 1, characterized in that: The method further comprises: Obtaining the POI distribution vectors of all transfer stations at each transfer station to form a POI distribution matrix for each transfer station; wherein the POI distribution vector of each station is composed of the number of POIs of various types in the buffer zone of each station; The ratio between the demand of each station transferred at each transfer station and the starting and ending point demand of the corresponding transfer station is used as the weight of each station transferred at each transfer station; The POI distribution matrix is ​​weighted by using the weights to obtain the POI distribution of each transfer station, and the transfer function characteristics of each transfer station are obtained according to the POI distribution.

9. The method for identifying spatiotemporal patterns of subway-bus combination travel based on two-step clustering according to claim 8, characterized in that: The acquiring of the transfer function characteristics of each transfer station according to the POI distribution specifically includes: in, represents the transfer function characteristics of the p-th POI type in the i-th transfer station buffer, represents the POI distribution of the pth POI type in the buffer zone of the i-th transfer station, N i represents the total number of transfer POIs at the i-th transfer station, N P Represents the sum of the number of all POI types in the study area, and N is the sum of the number of all POIs in the study area.

Images