Urban functional area identification method fused with area representation

Abstract

The invention provides an urban functional area identification method fusing regional representation, which comprises the following steps: extracting regional multi-period space interaction characteristics based on mobile phone positioning data and urban POI data and Node2vec, extracting POI characteristics by utilizing a GloVe model, and introducing a TF-IDF index to extract semantic characteristics of a region; and a multi-head attention mechanism is adopted to carry out regional feature fusion so as to carry out function category identification. When the characteristic relation is captured, the information sharing among the areas is realized, and the characteristic redundancy is reduced to the greatest extent, so that the accuracy of urban functional area identification is effectively improved.

Description

Technical Field

[0001] The invention belongs to the field of spatial information technology, and in particular relates to a method for identifying urban functional areas by integrating regional representation. Background Art

[0002] A city is an open, complex and huge system composed of various forms of functional areas. Urban functional areas not only carry various functions of social and economic development, but also, as the basic unit of urban population activities, provide specific functions for human activities and bear specific travel and activity needs. Accurate identification of urban functional areas can help decision makers better understand the functional structure of the city, thereby promoting the coupling of the human-land system and urban sustainable development. Human daily activities in urban areas and mobility between areas can potentially reflect the functional structure of the city. Therefore, exploring the interaction characteristics between urban areas and the semantic characteristics within the areas from the perspective of population movement, and learning through feature representation, can provide a new path for the identification of urban functional areas.

[0003] In terms of the identification of urban functional areas, traditional methods mainly use spatial morphological features such as spectrum, texture and radar scattering coefficient of remote sensing images, or use the socioeconomic characteristics reflected by POI to identify urban functional areas. At present, most studies have begun to use social perception big data such as social media, mobile phone positioning, and taxi trajectories to reveal urban functions from the perspective of crowd movement patterns, such as "Ma Shuhong, Zhang Junjie, Chen Xifang, et al. Identifying urban functional areas using taxi time series data [J]. Journal of Jilin University (Engineering Edition): 1-10." The time series vector of the region is constructed through trajectory data, and on this basis, the improved DTW algorithm and clustering algorithm are used to divide the urban functional attributes. "Xiao Rui, Guo Yuxiang, Li Xinghua. Dynamic extraction of urban plot activity semantics based on topic model [J]. Remote Sensing Technology and Applications, 2023, 38(3): 649-661." The taxi trajectory data is converted into travel mode text, and the topic model is used to discriminate the semantics of urban functional areas. Other scholars have identified urban functional areas by integrating POI data with social perception data, such as "Yang Zhenshan, Su Jinhua, Yang Hang, et al. Refined research on urban functional areas based on multi-source data - taking Beijing as an example [J]. Geographical Research, 2021, 40(2): 477-494." The category of urban functional areas is determined by weighted fusion of quantified POI category vectors and population heat. "Jing C, Hu Y, Zhang H, et al. Context-Aware Matrix Factorization for the Identification of Urban Functional Regions with POI and Taxi OD Data [J]. ISPRS International Journal of Geo-Information, 2022, 11(6): 351. DOI: 10.3390 / ijgi11060351" constructs a decomposition feature matrix of POI data and taxi OD trajectory data and derives a fusion matrix to perform data fusion and then divide urban functional areas by clustering.

[0004] The shortcomings of existing technologies are mainly reflected in the following aspects: (1) Most studies simply extract the temporal changes of human activities in trajectory data, or only use traffic characteristics to express the spatial interactivity of the region, so the characteristics of human activities reflected in trajectory data are not deeply mined; (2) When using POI data to obtain regional semantic features, global features and the public awareness of POI points are not taken into account, which may lead to misjudgment of functional area categories; (3) Some existing studies often ignore the similarities between different data when fusing multi-source data, resulting in feature redundancy, and it is difficult to capture the relationship between data, which needs further improvement. Summary of the invention

[0005] To this end, in view of the defects and shortcomings of the existing technology, the present invention aims to propose a method for identifying urban functional areas by integrating regional representation. Based on mobile phone positioning data and POI data, the method deeply mines the spatial interaction characteristics of multiple time periods of the region through the Node2vec algorithm, extracts POI features using the GloVe model, and introduces the TF-IDF index that can represent the public awareness of POI to extract the semantic features of the region. Finally, a multi-head attention mechanism is used to achieve effective fusion of regional features, while capturing feature relationships, realizing information sharing between regions, minimizing feature redundancy, and effectively improving the accuracy of urban functional area identification.

[0006] The technical solution specifically adopted by the present invention to solve the technical problem is:

[0007] A method for identifying urban functional areas by integrating regional representation: Based on mobile phone positioning data and urban POI data, Node2vec is used to extract regional multi-period spatial interaction features, the GloVe model is used to extract POI features, and the TF-IDF index is introduced to extract the semantic features of the region; a multi-head attention mechanism is used to fuse regional features for functional category recognition.

[0008] Furthermore, the extraction of regional multi-period spatial interaction features based on Node2vec includes:

[0009] Divide the study area: Use OSM data to divide the study area into road network areas;

[0010] Construct a weighted directed graph of user travel: Divide user travel into several specific time periods, then match user travel OD to road network areas, and use travel volume and travel distance to construct a weighted directed graph of user travel in specific time periods on weekdays and weekends;

[0011] The Node2vec model is used to obtain the representation vector of each road network area, thereby obtaining the multi-period spatial interaction characteristics of the urban area.

[0012] Furthermore, the extraction of regional multi-period spatial interaction features based on Node2vec is specifically as follows:

[0013] The activities of urban people are mapped into a travel directed graph in the mobile network space. The crowd travel network is constructed with urban road network areas as nodes and the flow of urban people between areas as edges. The Node2vec algorithm is used to obtain the spatial interaction characteristics of different areas at different time periods:

[0014] Divide user travel into multiple specific time periods; then match user travel OD to the road network area, and construct a multi-time user travel directed graph set {G1, G2, G3, G4, G5, G6}; the user travel directed graph of time period t is defined as: G t =(V t ,E t ), where V t represents the set of nodes in the tth period, E t represents the directed edge set of the graph below in the tth period; E t Defined as: Among them, R i ,R j ∈V t , respectively represent the starting and ending nodes connected by the edge, represents edge weight;

[0015] Each urban road network area is a node in the travel directed graph. The greater the travel volume between two areas, the stronger the connection, which is measured by the edge weight;

[0016] In the t period, from node R i To Node R j Edge weight Determined by formula (1):

[0017]

[0018] in, Indicates that in the tth period, from node R i Arrival at Node R j The amount of travel, Indicates that the number of nodes from R i Arrival at Node R j The average distance of all trips is Indicates that within this period, all nodes in the graph are reached to node R j The total amount of trips;

[0019] Based on the constructed weighted directed graph of user travel, Node2vec is introduced to perform a strategic random walk on the graph to simulate the movement trajectory of people in various areas of the city: Assuming that the current user was at node s in the previous step and is currently at node i, then the transition probability P(v j |v i ) is calculated as follows:

[0020]

[0021] Where Z is the normalization operator, π i,j is the non-normalized transition probability, and the calculation formula is as follows:

[0022] π i,j =α pq (s,j) * w i,j (3)

[0023] Where w i,j The edge weights between nodes calculated by formula (1) are: pq (s,j) is calculated by formula (4):

[0024]

[0025] where d s,j It refers to the shortest path between node s and node j. Parameters p and q respectively control the direction of wandering, whether it is walking towards the area that has been visited or walking towards the area that has not been visited. When node s coincides with node j, the shortest path is 0, and the return parameter p controls the direction inward. When node s is directly connected to node j, the shortest path is 1. When node s and node j require two hops to connect, the shortest path is 2, and the entry and exit parameter q controls the wandering direction outward.

[0026] Furthermore, the method of extracting POI features using the GloVe model and introducing the TF-IDF index to extract the semantic features of the region includes: constructing a POI buffer document of the shortest path sequence, and constructing a city function corpus on this basis; training the city function corpus using the GloVe model to obtain a POI category feature vector; introducing the TF-IDF weight calculation method to evaluate the importance of each POI category in the region, and calculating the representation vector of the region, that is, the semantic features of the region.

[0027] Furthermore, the method of extracting POI features using the GloVe model and introducing the TF-IDF index to extract the semantic features of the region is specifically as follows:

[0028] Firstly, the POI buffer zone document based on the shortest path sequence is used to construct the urban function corpus; the specific steps are as follows: construct a buffer zone with each POI as the center, calculate the distance between all POI points in the buffer zone, and select the two POIs with the farthest distance as the starting point S p and end point D p , construct a collection sequence {S p ,D p}; Then put the remaining POI points in the buffer into the set to be inserted W = {P1, P2, P3, ..., P n}, use the greedy algorithm to insert the POIs in W into the set sequence one by one, ensuring that the path distance formed after insertion is the shortest; repeat this process until the set to be inserted is empty;

[0029] Then, the GloVe model is used to train the city corpus. In this process, the POI buffer is regarded as a document, and the category name of the third-level classification of POI in the buffer is regarded as the word in the document. The sentence in the document is composed of these words in the shortest path sequence. The representation vector of each POI category obtained by training can effectively reflect the semantic information of the region.

[0030] POIs are reclassified into several POI types, and TF-IDF weights are introduced to calculate the importance of different POI types; the final calculation formula for the region representation vector is as follows:

[0031]

[0032] in, Represents the feature vector of the i-th region, type vec (P i,j ) represents the representation vector of POI category j in the i-th region, TF-IDF i,j It represents the TF-IDF weight value corresponding to the POI type of category j in the i-th region; N represents the number of POIs in the region.

[0033] Furthermore, the use of a multi-head attention mechanism to fuse regional features for functional category identification includes: inputting spatial interactions and semantic features of each time period into the model, calculating weights through the multi-head attention mechanism, and generating a final feature representation; identifying urban functional areas: the model outputs the functional type probability of each area, and selects the highest probability as the identification result.

[0034] Furthermore, the multi-head attention mechanism is used to fuse regional features for functional category recognition as follows:

[0035] The spatial interaction features and semantic features of the regions in each time period are used as the input of the model, and the model is trained using the labeled data of the functional area samples. Through the multi-head attention mechanism, these features are integrated into a comprehensive regional representation vector to obtain the recognition results of urban functional areas:

[0036] Input various types of spatial interaction features and semantic features of the region;

[0037] The input vector is subjected to a series of linear transformations through the encoding layer to obtain the Query value, Key value, and Value value of the multi-head attention mechanism;

[0038] The weight of each input feature vector is calculated independently by each head and multiplied by the corresponding Value value to obtain the output of multiple attention heads; the final feature representation is formed after splicing and linear transformation;

[0039] The probability of each region belonging to each functional area category is calculated through the fully connected layer FCN and the Softmax function; when the probability of a certain category is the highest and higher than 50%, the function of the region belongs to this category; if the probabilities of all categories are lower than 50%, the region is marked as a mixed functional area.

[0040] Furthermore, it also includes data preprocessing, specifically:

[0041] Collect and pre-process mobile phone positioning data and city POI data:

[0042] Mobile phone positioning data preprocessing: Eliminate missing data, duplicate data, invalid data, redundant data and data outside the study area; Eliminate data with ping-pong switching and drift point data; Eliminate data with less than 10 records per day;

[0043] Preprocessing of urban POI data: Eliminate data records outside the study area and duplicate data records; Convert POI coordinate system to WGS84 coordinate system;

[0044] Identify stop points and extract trip OD data:

[0045] Stay point identification: The ST-DBSCAN algorithm is used to mark the stay cluster data and remove noise data. The earliest time and the latest time in the same cluster of the same user are used as the start time and end time of the stay point, respectively. The geometric center of the point in the cluster is calculated to obtain the longitude and latitude of the stay point. Records with a stay time greater than 10 minutes are selected as valid records, and records with 1 stay point and more than 10 stay points are removed.

[0046] Trip OD extraction: A trajectory between two consecutive stop points of the same user is regarded as a trip OD; OD pairs with a travel distance of less than 2 km and a travel time of less than 15 minutes are eliminated to finally obtain valid travel OD data.

[0047] And, an urban functional area identification system integrating regional representation, comprising:

[0048] Multi-period spatial interaction feature extraction module, used to extract regional multi-period spatial interaction features based on Node2vec;

[0049] The semantic feature extraction module is used to extract POI features using the GloVe model and introduce the TF-IDF index to extract the semantic features of the region;

[0050] And a recognition module, which uses a multi-head attention mechanism to fuse regional features for functional category recognition.

[0051] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the urban functional area identification method using fused area representation as described above are implemented.

[0052] Compared with the prior art, the present invention and its preferred embodiment have at least the following outstanding advantages: (1) The travel volume and travel distance are introduced to construct a weighted directed graph of user travel in six different time periods, and based on this, the Node2vec algorithm is trained to effectively capture the dynamic connection between different areas in different time periods, and at the same time obtain a distinctive regional representation vector, which provides rich feature information for the subsequent identification of functional areas. (2) After using the GloVe model to obtain the POI category feature vector, the TF-IDF weight calculation method is introduced to evaluate the importance of each POI category in the region, which can extract more representative regional semantic feature vectors, thereby more accurately reflecting the actual impact of different POIs in the region and improving the accuracy of functional area identification. (3) The proposed fusion method can make full use of the complementarity of different types of data, has higher classification accuracy, and has a higher improvement in the recognition accuracy of residential areas and mixed areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0054] Figure 1 This is a technical framework diagram of urban functional area identification fused with regional representation according to an embodiment of the present invention;

[0055] Figure 2 This is an example diagram of a directed graph of a trip according to an embodiment of the present invention;

[0056] Figure 3 A schematic diagram of a random walk strategy according to an embodiment of the present invention;

[0057] Figure 4 This is a flow chart of regional feature fusion and urban functional area identification based on a multi-head attention mechanism in an embodiment of the present invention;

[0058] Figure 5 This is a recognition result diagram of the urban functional area according to an embodiment of the present invention;

[0059] Figure 6 1 is the confusion matrix of four comparative experiments in the embodiments of the present invention. DETAILED DESCRIPTION

[0060] In order to make the features and advantages of this patent more obvious and easy to understand, the following embodiments are specifically described in detail as follows:

[0061] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0062] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0063] Based on mobile phone positioning data, city POI data and OSM road network data, the embodiment of the present invention uses the Node2vec algorithm and the GloVe model to extract the multi-time spatial interaction features and semantic features of the region respectively, and uses the multi-head attention mechanism to fully integrate these features for functional category recognition. Its main implementation process includes:

[0064] (1) Collect and preprocess mobile phone positioning data and city POI data;

[0065] (2) Identify stop points and extract trip OD data;

[0066] (3) Extracting regional multi-period spatial interaction features based on Node2vec;

[0067] (4) Extracting regional semantic features based on GloVe;

[0068] (5) Identification of urban functional areas by integrating multi-head attention mechanism;

[0069] The technical solution of the embodiment of the present invention is described in detail below.

[0070] The purpose of this invention is to propose a method for identifying urban functional areas by integrating regional representation. The basic framework of the model is as follows Figure 1 As shown, the following steps are included:

[0071] The first step is to collect and preprocess mobile phone positioning data and city POI data;

[0072] Mobile phone positioning data preprocessing: Eliminate missing data, duplicate data, invalid data, redundant data and data outside the study area; Eliminate data with ping-pong switching and drift point data; Eliminate data with too few records per day (less than 10). The single trajectory data is shown in Table 1. (2) Urban POI data preprocessing: Eliminate data records outside the study area and duplicate data records; Convert the POI coordinate system to the WGS84 coordinate system.

[0073] Table 1 Single track data example

[0074]

[0075]

[0076] In the second step, the stay points are identified and the trip OD data are extracted;

[0077] (1) Stay point identification: The ST-DBSCAN algorithm is used to mark the stay cluster data and remove noise data. The earliest time and the latest time in the same cluster of the same user are used as the start and end time of the stay point, respectively. The geometric center of the point in the cluster is calculated to obtain the longitude and latitude of the stay point; the records with a stay time greater than 10 minutes are selected as valid records, and the number of stay points is 1 and too many stay points (greater than 10) are removed. (2) Travel OD extraction: The trajectory between two consecutive stay points of the same user is regarded as a travel OD. OD pairs with a travel distance that is too short (less than 2 km) and a travel time that is too short (less than 15 minutes) are removed, and finally valid travel OD data is obtained.

[0078] The third step is to extract the regional multi-period spatial interaction features based on Node2vec;

[0079] (1) Divide the study area: Use OSM data to divide the study area into road network areas, which serve as urban unit areas for subsequent experimental analysis; (2) Construct a weighted directed graph of user travel: Divide user travel into six specific time periods, then match user travel OD to road network areas, introduce travel volume and travel distance to construct a weighted directed graph of user travel in six specific time periods on weekdays and weekends; (3) Use the Node2vec model to obtain the representation vector of each road network area, thereby obtaining the multi-time spatial interaction characteristics of the urban area.

[0080] The fourth step is to extract regional semantic features based on GloVe;

[0081] (1) Construct a POI buffer zone document of the shortest path sequence, and on this basis construct an urban function corpus; (2) Use the GloVe model to train the urban function corpus and obtain the POI category feature vector; (3) Introduce the TF-IDF weight calculation method to evaluate the importance of each POI category in the region and calculate the representation vector of the region, that is, the semantic feature of the region.

[0082] Step 5: Identification of urban functional areas by integrating multi-head attention mechanism;

[0083] (1) Fusion of regional features: The spatial interactions and semantic features of each time period are input into the model, and the weights are calculated through the multi-head attention mechanism to generate the final feature representation; (2) Identification of urban functional areas: The model outputs the functional type probability of each area and selects the highest probability as the recognition result.

[0084] In the above steps, the extraction of regional multi-period spatial interaction features based on Node2vec, the extraction of regional semantic features based on GloVe, and the identification of urban functional areas by integrating the multi-head attention mechanism are the key points of this invention, which will be discussed in detail in the following subsections. (1) Extraction of regional multi-period spatial interaction features based on Node2vec.

[0085] The activities of urban people in real physical space can be mapped into a travel directed graph in mobile network space. Based on this, the present invention constructs a crowd travel network with urban road network areas as nodes and the flow of urban people between areas as edges, and uses the Node2vec algorithm to explore the spatial interaction characteristics of different areas in different time periods.

[0086] (a) Constructing a weighted directed graph of user travel

[0087] The user travel is divided into six specific time periods, namely: morning peak (7:00-9:00), morning working time (9:00-11:00), lunch time (11:00-13:00), afternoon working time (13:00-17:00), evening peak (17:00-19:00) and evening entertainment time (19:00-23:00). Then the user travel OD is matched to the road network area (node) to construct a multi-time user travel directed graph set {G1, G2, G3, G4, G5, G6}. The user travel directed graph for time period t is defined as: G t =(V t ,E t ), where V t represents the set of nodes in the tth period, and E t represents the directed edge set of the graph below during this period. t Defined as: Among them, R i ,R j ∈V t , respectively represent the starting and ending nodes connected by the edge, Represents edge weight.

[0088] Each urban road network area is a node in the travel directed graph, with a one-way arrow and a number indicating the direction and number of user trips between nodes. The thicker the arrow line, the greater the travel volume between the two areas, and the stronger the connection. The strength of this connection is generally measured by the weight of the edge. Figure 2shown.

[0089] Considering that most people have fixed travel routes and prefer to choose shorter routes, the tth period starts from node R i To Node R j Edge weight It can be determined by formula (1):

[0090]

[0091] in, Indicates that in the tth period, from node R i Arrival at Node R j The amount of travel, Indicates that the number of nodes from R i Arrival node Rj 的 The average distance of all trips, Indicates that within this period, all nodes in the graph are reached to node R j The total amount of trips.

[0092] (b) Region vector representation based on Node2vec

[0093] Based on the constructed user travel weighted directed graph, Node2vec is introduced to perform strategic random walks on the graph to simulate the movement trajectory of people in various areas of the city. The random walk strategy is as follows: Figure 3 Assume that the user was at node s in the previous step and is currently at node i, then the probability of transitioning to node j in the next step is P(v j |v i ) is calculated as follows:

[0094]

[0095] Where Z is the normalization operator, π i,j is the non-normalized transition probability, and the calculation formula is as follows:

[0096] π i,j =α pq (s,j) * w i,j (3)

[0097] Where w i,j The edge weights between nodes calculated by formula (1) are: pq (s,j) is calculated by formula (4):

[0098]

[0099] where d s,jIt refers to the shortest path between node s and node j. Parameters p and q control the direction of the walk, whether it is to walk to the area that has been visited or to walk to the area that has not been visited. When node s and node j overlap, the shortest path is 0, and the return parameter p controls the direction inward; when node s and node j are directly connected, the shortest path is 1; when node s and node j need two hops to connect, the shortest path is 2, and the entry and exit parameter q controls the direction of the walk outward.

[0100] (2) Extract regional semantic features based on GloVe.

[0101] Human activities usually revolve around POIs. The distribution of POIs in cities contains a lot of semantic information. Based on this, the present invention uses the GloVe model to explore the semantic features of different areas in the city.

[0102] (a) Constructing an urban function corpus

[0103] In order to reflect the distribution characteristics of POIs in geographic space and maintain the positional relationship between adjacent POIs, this paper uses POI buffer documents based on the shortest path sequence to construct the urban function corpus. The specific steps are as follows: construct a buffer with each POI as the center, calculate the distance between all POI points in the buffer, and select the two POIs with the farthest distance as the starting point S p and end point D p , construct a collection sequence {S p ,D p Then put the remaining POI points in the buffer into the set to be inserted W = {P1, P2, P3, ..., P n}, the greedy algorithm is used to insert the POIs in W into the set sequence one by one, ensuring that the path distance formed after insertion is the shortest. Repeat this process until the set to be inserted is empty.

[0104] (b) POI feature vector representation based on GloVe

[0105] The GloVe model is used to train the city corpus. In this process, the POI buffer is regarded as a document, the category of the POI in the buffer (the category name of the third-level classification of POI) is regarded as the word in the document, and the sentence in the document is composed of these words in the shortest path sequence. The representation vector of each POI category obtained by training can effectively reflect the semantic information of the region and provide support for the subsequent calculation of the semantic features of the region.

[0106] (c) Calculate the region representation vector

[0107] In order to reduce the impact of the difference in the number of POIs of different categories, the present invention reclassifies POIs into cognitive type and density type, as shown in Table 2, with a total of 16 POI types, and introduces TF-IDF weights to calculate the importance of different POI types. The final calculation formula for the region representation vector is as follows:

[0108]

[0109] in, Represents the feature vector of the i-th region, type vec (P i,j ) represents the representation vector of POI category j in the i-th region, TF-IDF i,j Indicates the TF-IDF weight value corresponding to the POI type of category j in the i-th region. N indicates the number of POIs in the region.

[0110] Table 2 POI cognitive density classification system

[0111]

[0112] (3) Identification of urban functional areas by integrating multi-head attention mechanism

[0113] The spatial interaction features and semantic features of the regions in each time period are used as the input of the model, and the model is trained using the labeled data of the functional area samples. Through the multi-head attention mechanism, these features are integrated into a comprehensive regional representation vector to obtain the recognition results of urban functional areas. The specific workflow is as follows Figure 4 shown.

[0114] (a) Input 12 types of spatial interaction features (6 each for weekdays and weekends) and the semantic features of the region.

[0115] (b) A series of linear transformations are performed on the input vector through the encoding layer to obtain the Query value, Key value, and Value value of the multi-head attention mechanism.

[0116] (c) Each head independently calculates the weight of each input feature vector and multiplies it with the corresponding Value value to obtain the output of multiple attention heads. After splicing and linear transformation, the final feature representation is formed.

[0117] (d) The probability of each region belonging to each functional area category is calculated through the fully connected layer (FCN) and the Softmax function. When the probability of a category is the highest and higher than 50%, the region function belongs to this category; if the probability of all categories is lower than 50%, the region is marked as a mixed functional area.

[0118] The efficacy of the solution provided by the present invention is further demonstrated and illustrated by means of specific test examples below:

[0119] Based on the mobile phone positioning data and POI data of a large city in China, the present invention adopts the method of integrating regional representation to identify the urban functional areas. The identification results of urban functional areas are as follows: Figure 5 As shown, we finally get six types of functional areas: residential, industrial, commercial, public administration and public services, leisure and sightseeing, and mixed. Figure 5 As shown in the classification results, 30 areas of each functional area were randomly selected for inspection, and the classification results were compared with the real functional area categories obtained from Amap to verify the accuracy of the classification results. The results of the confusion matrix are shown in Table 3. The overall accuracy OA is 81%, and the value of the Kappa coefficient is 0.77, indicating that the recognition results are highly consistent with the real regional functions.

[0120] Table 3 Classification confusion matrix

[0121]

[0122]

[0123] In order to further verify the effectiveness of the method proposed in the present invention, two commonly used methods in the field were selected for comparative analysis, and the corresponding accuracy results of each method are shown in Table 4. The results show that the kappa coefficient and OA value of the method of the present invention are higher than those of the method based on DTW_KNN and the method based on Word2Vec. The DTW_KNN method mainly relies on the time series vector to express the interaction between regions, but ignores the global influence of the region. In the end, the functional area division result obtained by clustering traffic features alone is difficult to distinguish regions with complex functional structures. Although the Word2Vec method takes into account the quantitative characteristics and spatial distribution of POIs, and represents the semantic characteristics of different regions through high-dimensional vectors, it obtains a better classification effect than DTW_KNN, but because it does not consider the public's awareness of POIs and global characteristics, the classification results are still not accurate enough. The method proposed in the present invention improves the above shortcomings. When exploring spatial interactivity, it not only considers the global influence, but also expresses the local characteristics of the region; in the semantic feature analysis, it takes into account the global distribution structure of POIs and the public's awareness, so a more accurate classification result is obtained.

[0124] Table 4 Comparison of accuracy of different methods

[0125] method Kappa OA (%) DTW_KNN 0.41 0.51 Word2Vec 0.51 0.60 Method of the present invention 0.77 0.81

[0126] Since the method proposed in this invention consists of multiple parts, ablation experiments were conducted to verify the influence of different data on functional area recognition and the effect of feature fusion methods. The first type of experiment used single-source data for functional area recognition, using a method based on the Node2vec algorithm and mobile phone positioning data (method 1), and a method based on the GloVe model and POI data (method 2). The second type of experiment used the traditional weighted fusion method instead of the multi-head attention mechanism to recognize urban functional areas (method 3). The confusion matrix of the four method comparison experiments is shown in Figure 2. Figure 6 As shown, the results show that the recognition accuracy of the method of the present invention is the highest, which is 81.11%, followed by method 2 (75%), method 3 (67.78%), and method 1 (63.33%) has the lowest accuracy. Among them, method 2 performs well in identifying industrial areas and leisure and sightseeing areas, and only performs poorly in identifying residential areas and mixed areas, but is still better than method 1 and method 3. Compared with Word2Vec, which also uses word embedding technology, this method also performs better; except for the good performance of commercial areas and leisure and sightseeing areas, method 1 has relatively low recognition accuracy in the remaining categories, especially poor performance in the recognition of residential areas and mixed areas, but the recognition accuracy is still higher than the DTW_KNN algorithm; method 3 uses two kinds of data, and compared with method 1, it has a significant improvement in the recognition of industrial areas and mixed areas, and a slight improvement in the recognition of residential areas, but the recognition accuracy of commercial areas is reduced, and the recognition accuracy is not as good as method 2 which uses one kind of data. Compared with the above three comparison methods, the method of the present invention has a significant improvement in the recognition accuracy of residential areas and mixed areas, and also has a good effect in the recognition of other categories.

[0127] The above experiments show that each link of feature extraction in the method of the present invention is very necessary, and the effect of fully integrating the spatial interaction features of multi-scale regions and the semantic features of regions for urban functional area recognition is better than a single data source and a simple fusion method. By fully integrating the two types of data, the shortcomings of single data can be compensated and higher recognition accuracy can be achieved.

[0128] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0129] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0130] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0131] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0132] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0133] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

[0134] This patent is not limited to the above-mentioned optimal implementation mode. Anyone can derive other forms of a method for identifying urban functional areas with integrated regional representation under the inspiration of this patent. All equal changes and modifications made according to the scope of the patent application of the present invention should be covered by this patent.

Claims

1. A method for identifying urban functional areas by integrating regional representation, characterized in that: Based on mobile phone positioning data and urban POI data, Node2vec is used to extract regional multi-period spatial interaction features, the GloVe model is used to extract POI features and the TF-IDF index is introduced to extract the semantic features of the region; the multi-head attention mechanism is used to fuse regional features for functional category recognition.

2. The method for identifying urban functional areas by integrating regional representation according to claim 1, characterized in that: The method of extracting regional multi-period spatial interaction features based on Node2vec includes: Divide the study area: Use OSM data to divide the study area into road network areas; Construct a weighted directed graph of user travel: Divide user travel into several specific time periods, then match user travel OD to road network areas, and use travel volume and travel distance to construct a weighted directed graph of user travel in specific time periods on weekdays and weekends; The Node2vec model is used to obtain the representation vector of each road network area, thereby obtaining the multi-period spatial interaction characteristics of the urban area.

3. The method for identifying urban functional areas by integrating regional representation according to claim 2, characterized in that: The method of extracting regional multi-period spatial interaction features based on Node2vec is specifically as follows: The activities of urban people are mapped into a travel directed graph in the mobile network space. The crowd travel network is constructed with urban road network areas as nodes and the flow of urban people between areas as edges. The Node2vec algorithm is used to obtain the spatial interaction characteristics of different areas at different time periods: Divide user travel into multiple specific time periods; then match user travel OD to the road network area, and construct a multi-time user travel directed graph set {G1, G2, G3, G4, G5, G6}; the user travel directed graph of time period t is defined as: G t =(V t ,E t ), where V t represents the set of nodes in the tth period, E t represents the directed edge set of the graph below in the tth period; E t Defined as: Among them, R i ,R j ∈V t , respectively represent the starting and ending nodes connected by the edge, represents edge weight; Each urban road network area is a node in the travel directed graph. The greater the travel volume between two areas, the stronger the connection, which is measured by the edge weight; In the t period, from node R i To Node R j Edge weight Determined by formula (1): in, Indicates that in the tth period, from node R i Arrival at node R j The amount of travel, Indicates that the number of nodes from R i Arrival at node R j The average distance of all trips is Indicates that within this period, all nodes in the graph are reached to node R j The total amount of trips; Based on the constructed weighted directed graph of user travel, Node2vec is introduced to perform a strategic random walk on the graph to simulate the movement trajectory of people in various areas of the city: Assuming that the current user was at node s in the previous step and is currently at node i, then the transition probability p(v j |v i ) is calculated as follows: Where Z is the normalization operator, π i,j is the non-normalized transition probability, and the calculation formula is as follows: p i,j =a pq (s,j) * w i,j (3) Where w i,j The edge weights between nodes calculated by formula (1) are: pq (s,j) is calculated by formula (4): where d s,j It refers to the shortest path between node s and node j. Parameters p and q respectively control the direction of wandering, whether it is walking towards the area that has been visited or walking towards the area that has not been visited. When node s coincides with node j, the shortest path is 0, and the return parameter p controls the direction inward. When node s is directly connected to node j, the shortest path is 1. When node s and node j require two hops to connect, the shortest path is 2, and the entry and exit parameter q controls the wandering direction outward.

4. The method for identifying urban functional areas by integrating regional representation according to claim 1, characterized in that: The method of extracting POI features using the GloVe model and introducing the TF-IDF index to extract the semantic features of the region includes: constructing a POI buffer document of the shortest path sequence, and constructing an urban function corpus on this basis; training the urban function corpus using the GloVe model to obtain a POI category feature vector; introducing a TF-IDF weight calculation method to evaluate the importance of each POI category in the region, and calculating a representation vector of the region, that is, the semantic features of the region.

5. The method for identifying urban functional areas by integrating regional representation according to claim 4, characterized in that: The method of using the GloVe model to extract POI features and introducing the TF-IDF index to extract the semantic features of the region is specifically as follows: Firstly, the POI buffer zone document based on the shortest path sequence is used to construct the urban function corpus; the specific steps are as follows: construct a buffer zone with each POI as the center, calculate the distance between all POI points in the buffer zone, and select the two POIs with the farthest distance as the starting point S p and end point D p , construct a collection sequence {S p ,D p }; Then put the remaining POI points in the buffer into the set to be inserted W = {P1, P2, P3, ..., P n }, use the greedy algorithm to insert the POIs in W into the set sequence one by one, ensuring that the path distance formed after insertion is the shortest; repeat this process until the set to be inserted is empty; Then, the GloVe model is used to train the city corpus. In this process, the POI buffer is regarded as a document, and the category name of the third-level classification of POI in the buffer is regarded as the word in the document. The sentence in the document is composed of these words in the shortest path sequence. The representation vector of each POI category obtained by training can effectively reflect the semantic information of the region. POIs are reclassified into several POI types, and TF-IDF weights are introduced to calculate the importance of different POI types; the final calculation formula for the region representation vector is as follows: in, Represents the feature vector of the i-th region, type vec (P i,j ) represents the representation vector of POI category j in the i-th region, TF-IDF i,j It represents the TF-IDF weight value corresponding to the POI type of category j in the i-th region; N represents the number of POIs in the region.

6. The method for identifying urban functional areas by integrating regional representation according to claim 1, characterized in that: The use of a multi-head attention mechanism to fuse regional features for functional category recognition includes: inputting spatial interactions and semantic features of each time period into the model, calculating weights through the multi-head attention mechanism, and generating a final feature representation; identifying urban functional areas: the model outputs the functional type probability of each area, and selects the highest probability as the recognition result.

7. The method for identifying urban functional areas by integrating regional representation according to claim 6, characterized in that: The multi-head attention mechanism is used to fuse regional features for functional category recognition as follows: The spatial interaction features and semantic features of the regions in each time period are used as the input of the model, and the model is trained using the labeled data of the functional area samples. Through the multi-head attention mechanism, these features are integrated into a comprehensive regional representation vector to obtain the recognition results of urban functional areas: Input various types of spatial interaction features and semantic features of the region; The input vector is subjected to a series of linear transformations through the encoding layer to obtain the Query value, Key value, and Value value of the multi-head attention mechanism; The weight of each input feature vector is calculated independently by each head and multiplied by the corresponding Value value to obtain the output of multiple attention heads; the final feature representation is formed after splicing and linear transformation; The probability of each region belonging to each functional area category is calculated through the fully connected layer FCN and the Softmax function; when the probability of a certain category is the highest and higher than 50%, the function of the region belongs to this category; if the probabilities of all categories are lower than 50%, the region is marked as a mixed functional area.

8. The method for identifying urban functional areas by integrating regional representation according to claim 1, characterized in that: It also includes data preprocessing, specifically: Collect and pre-process mobile phone positioning data and city POI data: Mobile phone positioning data preprocessing: Eliminate missing data, duplicate data, invalid data, redundant data and data outside the study area; Eliminate data with ping-pong switching and drift point data; Eliminate data with less than 10 records per day; Preprocessing of urban POI data: Eliminate data records outside the study area and duplicate data records; Convert POI coordinate system to WGS84 coordinate system; Identify stop points and extract trip OD data: Stay point identification: The ST-DBSCAN algorithm is used to mark the stay cluster data and remove noise data. The earliest time and the latest time in the same cluster of the same user are used as the start time and end time of the stay point, respectively. The geometric center of the point in the cluster is calculated to obtain the longitude and latitude of the stay point. Records with a stay time greater than 10 minutes are selected as valid records, and records with 1 stay point and more than 10 stay points are removed. Trip OD extraction: A trajectory between two consecutive stop points of the same user is regarded as a trip OD; OD pairs with a travel distance of less than 2 km and a travel time of less than 15 minutes are eliminated to finally obtain valid travel OD data.

9. An urban functional area identification system integrating regional representation, characterized in that: include: Multi-period spatial interaction feature extraction module, used to extract regional multi-period spatial interaction features based on Node2vec; The semantic feature extraction module is used to extract POI features using the GloVe model and introduce the TF-IDF index to extract the semantic features of the region; And a recognition module, which uses a multi-head attention mechanism to fuse regional features for functional category recognition.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method for identifying urban functional areas represented by fusion areas as described in any one of claims 1 to 8 are implemented.

Images