A method and device for mining regional food culture preference based on interest points

Abstract

The present invention provides a method and device for mining regional food culture preferences based on points of interest, including presetting a number of regions and a number of cuisines, obtaining a preference index for each region for each cuisine, obtaining the amount of population mobility between two regions, constructing a dynamic weighted undirected graph data set, using regions as nodes, preference indexes as node features, population migration between two regions as edges, and the amount of population mobility as edge weights, and for the dynamic weighted undirected graph data set, using a community detection algorithm based on a dynamic graph attention network to extract node features, clustering several regions based on node features, and identifying regions with similar eating habits. The present invention uses a community detection algorithm based on a dynamic graph attention network to extract node features, clustering several regions based on node features, and can accurately identify regions with similar eating habits to explore the similarities in food culture between regions.

Description

Technical Field

[0001] The present invention relates to the technical field of data mining, and in particular to a method and device for mining regional food culture preferences based on points of interest. Background Art

[0002] Food has transcended its basic role of satisfying human physiological needs and has become an important medium for cultural inheritance. As an important part of human culture, food embodies unique regional characteristics. A variety of factors such as climate, economy and cultural customs have jointly created the dietary preferences of people in different regions. These preferences are not just personal choices, but also a broader cultural phenomenon, playing an important role in forming a unique food culture space. In addition, due to population migration and the integration of social culture, eating habits are increasingly transcending their original geographical boundaries. This cross-regional food migration promotes the reproduction of local social culture and reshapes individuals' cultural identity and their connection with the local cultural space.

[0003] Chinese cuisine is one of the richest and most diverse culinary traditions in the world, demonstrating how geography and culture merge to form unique eating habits. This article examines eight major cuisines in the Chinese culinary landscape: Shandong cuisine, Sichuan cuisine, Jiangsu cuisine, Cantonese cuisine, Fujian cuisine, Hunan cuisine, Anhui cuisine, and Zhejiang cuisine. Each cuisine has unique characteristics that deeply reflect the local environmental, climatic, and historical influences. For example, Shandong cuisine is known for seafood and clear soups, which is closely related to Shandong Province's coastal location and abundant seafood resources; while Sichuan cuisine is known for its bold and spicy flavors, which is closely related to Sichuan Province's humid climate and extensive cultivation of chili peppers. The uniqueness of each cuisine is not only reflected in ingredients and cooking techniques, but also reflects the profound influence of the natural environment, climatic conditions, and historical culture of their respective regions. Therefore, understanding regional dietary preferences is of great significance to the culinary industry, cultural historians, sociologists, and policymakers. This can guide restaurant site selection and menu design, promote the inheritance of cultural heritage, and support local economic development.

[0004] At present, some methods use map POI data to count the number of restaurants with different cuisines in different regions, but due to the different scales of restaurants in different regions and years, using absolute numbers cannot accurately reflect regional preferences. In addition, there are some methods that quantify regional preferences through reviews on review websites, but the large number of online water armies, fake reviews, and the complexity of review data processing reduce the feasibility and effectiveness of this method. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method and device for mining regional food culture preferences based on points of interest, which adopts a preference index to measure the popularity of each cuisine in each region, obtains the population flow and migration volume between two regions, establishes a dynamic weighted undirected graph data set, and applies an end-to-end graph clustering algorithm for community detection to explore the food culture preferences of various regions.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A method for mining regional food culture preferences based on points of interest, comprising:

[0008] Preset several regions and several cuisines, and obtain the preference index of each region for each cuisine;

[0009] Obtain the amount of population migration between two regions;

[0010] Construct a dynamic weighted undirected graph dataset, where regions are used as nodes, preference indexes are used as node features, population migration between regions is used as edges, and the amount of population migration is used as edge weights.

[0011] Graph clustering based on dynamic weighted undirected graph dataset: Use a community detection algorithm based on a dynamic graph attention network to extract node features, cluster several regions based on node features, and identify regions with similar eating habits.

[0012] Furthermore, the preference index of each region for each cuisine is obtained by dividing the number of restaurants of each cuisine in each region by the total number of restaurants in the region.

[0013] Furthermore, the population mobility between two regions is summed up to obtain the population mobility between each region.

[0014] Furthermore, in the process of constructing a dynamic weighted undirected graph data set, the first K edges whose population mobility and migration volume between two regions reaches a preset threshold are selected to establish an edge set.

[0015] Furthermore, a community detection algorithm based on a dynamic graph attention network is used to extract node features. The method for clustering several regions based on node features is as follows:

[0016] In the forward phase, a dynamic graph attention network is used to embed the region represented by each node, and the attention mechanism is used to extract the features of the nodes. Based on the features, the K-means clustering method is used to preliminarily cluster all regions, so that all regions are divided into multiple communities, and the modularity of each region is obtained;

[0017] In the backward phase, the embedding representation of the node is continuously updated by optimizing the modularity.

[0018] Furthermore, the method of using K-means clustering method to perform preliminary clustering of all regions based on features is as follows:

[0019] Each region represented by node j has a feature vector x j , each community k has a cluster center μ k , calculate the probability that the region represented by node j belongs to community k as r jk The formula is:

[0020]

[0021] Among them, l represents the index of all cluster centers, and β is a parameter that controls the hardness of cluster assignment;

[0022] Update cluster center μ k , the calculation formula is:

[0023]

[0024] Furthermore, the method of continuously updating the embedding representation of nodes by optimizing the modularity is:

[0025] Calculate the gradient propagation and define the function f i,l (μ,x) represents the relationship between the cluster center and the eigenvector, and the calculation formula is:

[0026]

[0027] in, is the cluster center μ i The lth component of ;

[0028] Use the implicit function theorem to calculate the partial derivative of the cluster center with respect to the eigenvector

[0029]

[0030] A device for mining regional food culture preferences based on points of interest, comprising:

[0031] The first acquisition module is used to preset a number of regions and a number of cuisines, and obtain the preference index of each region for each cuisine;

[0032] The second acquisition module is used to obtain the amount of population mobility between two regions;

[0033] Dynamic weighted undirected graph construction module, used to construct a dynamic weighted undirected graph dataset, where regions are used as nodes, preference indexes are used as node features, population migration between regions is used as edges, and the amount of population migration is used as edge weights;

[0034] Preference identification module, used for graph clustering based on dynamic weighted undirected graph datasets: node features are extracted using a community detection algorithm based on a dynamic graph attention network, several regions are clustered based on the node features, and regions with similar eating habits are identified.

[0035] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for mining regional food culture preferences based on points of interest when executing the program.

[0036] A non-transitory computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned method for mining regional food culture preferences based on points of interest.

[0037] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0038] The present invention provides a method for mining regional food culture preferences based on points of interest. The preference index is used to measure the popularity of each cuisine in each region, and the amount of population mobility and migration is used as an influencing factor. A dynamic weighted undirected graph dataset is constructed, and a community detection algorithm based on a dynamic graph attention network is used to extract node features. Based on the node features, several regions are clustered to identify regions with similar eating habits, so as to explore the similarities in food culture between regions. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0040] Figure 1 A flowchart of a method for mining regional food culture preferences based on points of interest according to the present invention;

[0041] Figure 2 is a schematic diagram of a graph clustering method of the present invention;

[0042] Figure 3 It is a result diagram of a specific implementation method of the present invention;

[0043] Figure 4 It is a schematic diagram of a device for mining regional food culture preferences based on points of interest according to the present invention. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0045] Example 1

[0046] Example 1 provides a method for mining regional food culture preferences based on interest points, such as Figure 1 As shown, including:

[0047] Step S1: Preset a number of regions and a number of cuisines, and obtain the preference index of each region for each cuisine;

[0048] Step S2: Obtain the amount of population migration between two regions;

[0049] Step S3: construct a dynamic weighted undirected graph dataset, in which regions are used as nodes, preference indexes are used as node features, population migration between regions is used as edges, and the amount of population migration is used as edge weights;

[0050] Step S4: Graph clustering based on a dynamic weighted undirected graph dataset: Use a community detection algorithm based on a dynamic graph attention network to extract node features, cluster several regions based on the node features, and identify regions with similar eating habits.

[0051] In the related technology, some methods currently use map POI data to count the number of restaurants with different cuisines in different regions. However, due to the different scales of restaurants in different regions and years, the use of absolute numbers cannot accurately reflect regional preferences. In addition, there are some methods that quantify regional preferences through comments on review websites, but a large number of online water armies, false comments, and the complexity of comment data processing reduce the feasibility and effectiveness of this method. This embodiment provides a method for mining regional food culture preferences based on points of interest, using a preference index to measure the popularity of each cuisine in each region, and using the amount of population mobility as an influencing factor to construct a dynamic weighted undirected graph data set, using a community detection algorithm based on a dynamic graph attention network to extract node features, clustering several regions based on node features, identifying regions with similar eating habits, and exploring the similarities in food culture between regions.

[0052] This embodiment can also explore the geographical distribution and spread of China's eight major cuisines. We use the preference index to measure the popularity of each cuisine in each province, taking the province as the region to form the smallest research unit, and deeply analyze the food culture characteristics of different regions. The study focuses on China's eight traditional cuisines - Anhui cuisine, Hunan cuisine, Fujian cuisine, Cantonese cuisine, Sichuan cuisine, Shandong cuisine, Jiangsu cuisine and Zhejiang cuisine. By analyzing the restaurant interest point data of each province, we explore the preferences of different regions for each cuisine, providing a new perspective for understanding the regional characteristics and diversity of Chinese food culture.

[0053] In step S1 of this embodiment, the regional food preference is first calculated from the POI data. The food culture preference index is used to represent the popularity of food in a certain region. The preference index of each region for each cuisine is obtained by dividing the number of restaurants of each cuisine in each region by the total number of restaurants in the region. The calculation formula is:

[0054]

[0055] Among them, r i represents the preference index of region i for a certain cuisine, S i represents the total number of restaurants in region i, n i Indicates the number of restaurants in region i that provide this cuisine.

[0056] This formula provides a simple and effective method to quantify the relative preference of different regions for different cuisines, which is convenient for further analysis and comparison. In terms of the geographical distribution and spread of the eight major cuisines in China, this formula can quantify the relative preference of each province for the eight major cuisines.

[0057] In step S2 of this embodiment, the population mobility between two regions is summed to obtain the population mobility between two regions, and the calculation formula is:

[0058] e i,j =p i,j +p j,i

[0059] Among them, e i,j represents the amount of population migration, p i,j represents the population migration from region i to region j, p j,i Represents the amount of population migration from region j to region i.

[0060] When exploring the geographical distribution and spread of the eight major Chinese cuisines, population mobility data were calculated to measure the strength of the connection between different provinces by analyzing the amount of population mobility between two provinces. For each pair of provinces, the calculation of population mobility data was based on the sum of the population mobility in both directions.

[0061] In step S3 of this embodiment, a dynamic weighted undirected graph dataset is constructed. According to step S1, the preference index of each province for the eight major Chinese cuisines is extracted from the restaurant POI data from 2012 to 2022.

[0062] For each region i, use an index vector x i =[PI1,PI2,...,PI D ] to represent the preference index of each cuisine in the region, where PI d represents the preference index for cuisine d, where D is the total number of cuisines.

[0063] In this example, in the study of the geographical distribution and spread of the eight major Chinese cuisines, each province is regarded as a node in the dynamic weighted undirected graph, and its preference index vector is used as the node feature. The edges in the graph represent the population migration between provinces. For each pair of provinces i and j, the weight of the edge is defined as the total migration volume e between the two provinces. i,j .

[0064] In step S3 of this embodiment, Figure 2 As shown in the figure, in the process of constructing a dynamic weighted undirected graph dataset, the data preprocessing step requires an edge selection strategy. Considering that less population migration usually means a lower probability of cuisine spread, a data preprocessing method based on "edge selection" is proposed. The core of this method is that for each region, only the relationship between the top K regions with the highest population migration volume is selected for analysis. In this way, the focus is focused on the connection between regions with higher migration rates to more accurately reflect the actual situation of cuisine spread.

[0065] Specifically, the first K edges whose population flow and migration between two regions reaches the preset threshold are selected to establish an edge set. The edge set can be expressed as:

[0066] E={(i,j)|j∈Top_K(i)}

[0067] Among them, E represents the edge set, and Top_K(i) represents the top K regions with the highest population mobility and migration volume to and from region i.

[0068] In step S4 of this embodiment, Figure 2As shown, graph clustering is performed. After data preprocessing, it is necessary to use a community detection algorithm to cluster the nodes to study the dietary similarities between regions. Traditional graph clustering models are usually only applicable to static graph structures and do not consider the weights of edges. On this basis, this embodiment has been improved by adopting a dynamic graph attention network (Temporal Graph Attention Network, T-GAT), using the attention mechanism as the core component of feature extraction. In this way, the model can effectively capture the dynamic changes of nodes and edges at different times, and can pay more attention to the weights and importance of edges more accurately, thereby significantly improving the performance of clustering.

[0069] Specifically, a dynamic graph attention model is established for learning, that is, a community detection algorithm based on a dynamic graph attention network is used to extract node features. The method for clustering several regions based on node features is as follows:

[0070] In the forward phase, a dynamic graph attention network is used to embed the region represented by each node, and the attention mechanism is used to extract the features of the nodes. Based on the features, the K-means clustering method is used to preliminarily cluster all regions, so that all regions are divided into multiple communities, and the modularity of each region is obtained;

[0071] In the backward phase, the embedding representation of the node is continuously updated by optimizing the modularity.

[0072] The method of using K-means clustering method to perform preliminary clustering of all regions based on features is as follows:

[0073] Each region represented by node j has a feature vector x j , each community k has a cluster center μ k , calculate the probability that the region represented by node j belongs to community k as r jk The formula is:

[0074]

[0075] Among them, l represents the index of all cluster centers, and β is a parameter that controls the hardness of cluster assignment;

[0076] Update cluster center μ k , the calculation formula is:

[0077]

[0078] The method of continuously updating the embedding representation of nodes by optimizing modularity is:

[0079] Calculate the gradient propagation and define the function f i,l (μ,x) represents the relationship between the cluster center and the eigenvector, and the calculation formula is:

[0080]

[0081] in, is the cluster center μ i The lth component of .

[0082] Use the implicit function theorem to calculate the partial derivative of the cluster center with respect to the eigenvector

[0083]

[0084] Since the exact back-propagation computational complexity is O(K 3 p 3 ), is not feasible in large-scale problems, so we use an approximate method. In this case, we observe that It is often dominated by its diagonal terms (such as the identity matrix), and this simplification greatly reduces the computational complexity, especially when the clusters are relatively balanced and well separated.

[0085] Through these two stages, the graph clustering algorithm can effectively cluster nodes and optimize parameters in back propagation to improve clustering results. Through this improved method, our model can not only process static graphs, but also dynamically adapt to changes in graph structure and attributes, better revealing complex relationships and patterns in the data.

[0086] In order to facilitate those skilled in the art to understand and implement the present invention, the present embodiment is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0087] In this specific implementation, a dataset consisting of 31 provinces, 8 cuisines, restaurant POIs and population flows from 2012 to 2022 is used, and the regional food culture preference mining based on points of interest provided in this embodiment is adopted:

[0088] For graph clustering, the dimensions of the hidden layer and output layer of the dynamic graph attention network are set to 50, and the number of attention heads is set to 4. After the dynamic graph attention network extracts node features, the K-means clustering algorithm is used to cluster the nodes into 5 categories. The clustering hardness coefficient β is set to 50. In addition, the community modularity after clustering is calculated as the loss function.

[0089] The province clustering results of the present invention are visualized and the clustering results of the nodes are marked with different colors.

[0090] exist Figure 3 In the figure, the node clusters of different colors are clearly separated, indicating that the present invention can effectively identify and distinguish different groups.

[0091] The present invention can cluster provinces according to the similarity of dietary styles. For example, Qinghai, Gansu, Xinjiang, Tibet, Sichuan, Shaanxi and Chongqing are classified as one cluster. These provinces generally prefer spicy and strong flavors, which are closely related to their geographical environment and climate characteristics. These regions are mainly beef and mutton, supplemented by strong seasonings such as peppers and cumin, forming distinct regional dietary characteristics. The yellow cluster includes northeastern provinces such as Liaoning, Jilin and Shandong. These regions have similar climates and are greatly influenced by Shandong cuisine. The dishes have a refreshing taste and focus on the original flavor and mellowness of the ingredients. In addition, Beijing, as a separate cluster, fully reflects its special status as the capital. Due to the high concentration of population mobility across the country, Beijing has formed a diversified food culture that is inclusive and not limited to a single cuisine. This result highlights the advantages of the regional food culture preference mining method based on points of interest provided in this embodiment in capturing the diversity and uniqueness of regional food culture.

[0092] This embodiment provides a method for mining regional food culture preferences based on points of interest, which identifies provincial groups with similar eating habits by applying graph clustering technology to the food preference data between provinces. In the sample analysis, a dynamic graph is constructed using multi-source data such as population migration and cuisine preference index, and the analysis process is optimized through clustering results and similarity measurement between nodes, thereby reducing data processing costs and achieving more accurate regional food culture preference analysis.

[0093] Example 2

[0094] Embodiment 2 provides a regional food culture preference mining device based on interest points, such as Figure 4 As shown, including:

[0095] The first acquisition module is used to preset a number of regions and a number of cuisines, and obtain the preference index of each region for each cuisine;

[0096] The second acquisition module is used to obtain the amount of population mobility between two regions;

[0097] Dynamic weighted undirected graph construction module, used to construct a dynamic weighted undirected graph dataset, where regions are used as nodes, preference indexes are used as node features, population migration between regions is used as edges, and the amount of population migration is used as edge weights;

[0098] The preference identification module is used for graph clustering. For the dynamic weighted undirected graph dataset, a community detection algorithm based on a dynamic graph attention network is used to extract node features, and several regions are clustered based on the node features to identify regions with similar eating habits.

[0099] Example 3

[0100] Embodiment 3 provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for mining regional food culture preferences based on points of interest when executing the program.

[0101] Example 4

[0102] Embodiment 4 provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for mining regional food culture preferences based on points of interest is implemented.

[0103] The memory in the embodiment of the present invention is used to store various types of data to support the operation of the electronic device. Examples of such data include: any computer program used to operate on the electronic device.

[0104] The method for mining regional food culture preferences based on points of interest disclosed in the embodiment of the present invention can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the method for mining regional food culture preferences based on points of interest can be completed by hardware integrated logic circuits or software instructions in the processor. The above-mentioned processor may be a general-purpose processor, a digital signal processor (DSP, Digital Signal Processor), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor can implement or execute the methods, steps and logic block diagrams disclosed in the embodiment of the present invention. The general-purpose processor may be a microprocessor or any conventional processor, etc. In combination with the steps of the method disclosed in the embodiment of the present invention, it can be directly embodied as a hardware decoding processor to execute, or it can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium, which is located in a memory, and the processor reads the information in the memory, and combines its hardware to complete the steps of the method for mining regional food culture preferences based on points of interest provided in the embodiment of the present invention.

[0105] In an exemplary embodiment, the electronic device may be implemented by one or more application specific integrated circuits (ASIC), DSP, programmable logic device (PLD), complex programmable logic device (CPLD), FPGA, general purpose processor, controller, microcontroller (MCU), microprocessor, or other electronic components to execute the aforementioned method.

[0106] It can be understood that the memory can be a volatile memory or a non-volatile memory, and can also include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), synchronous static random access memory (SSRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct rambus random access memory (DRRAM).The memories described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memories.

[0107] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for mining regional food culture preferences based on points of interest, characterized in that: include: Preset several regions and several cuisines, and obtain the preference index of each region for each cuisine; Obtain the amount of population migration between two regions; Construct a dynamic weighted undirected graph dataset, where regions are used as nodes, preference indexes are used as node features, population migration between regions is used as edges, and the amount of population migration is used as edge weights. Graph clustering based on dynamic weighted undirected graph dataset: Use a community detection algorithm based on a dynamic graph attention network to extract node features, cluster several regions based on node features, and identify regions with similar eating habits.

2. The method for mining regional food culture preferences based on points of interest according to claim 1, characterized in that: The preference index of each region for each cuisine is obtained by dividing the number of restaurants of each cuisine in each region by the total number of restaurants in the region.

3. The method for mining regional food culture preferences based on points of interest according to claim 1, characterized in that: The population mobility between two regions is obtained by adding up the population mobility between two regions.

4. The method for mining regional food culture preferences based on points of interest according to claim 1, characterized in that: In the process of constructing a dynamic weighted undirected graph dataset, the first K edges whose population mobility and migration volume between two regions reaches a preset threshold are selected to establish an edge set.

5. The method for mining regional food culture preferences based on points of interest according to claim 1, characterized in that: The community detection algorithm based on dynamic graph attention network is used to extract node features. The method of clustering several regions based on node features is as follows: In the forward phase, a dynamic graph attention network is used to embed the region represented by each node, and the attention mechanism is used to extract the features of the nodes. Based on the features, the K-means clustering method is used to preliminarily cluster all regions, so that all regions are divided into multiple communities, and the modularity of each region is obtained; In the backward phase, the embedding representation of the node is continuously updated by optimizing the modularity.

6. The method for mining regional food culture preferences based on interest points according to claim 5 is characterized in that: The method of using K-means clustering method to perform preliminary clustering of all regions based on features is as follows: Each region represented by node j has a feature vector x j , each community k has a cluster center μ k , calculate the probability that the region represented by node j belongs to community k as r jk The formula is: Among them, l represents the index of all cluster centers, and β is a parameter that controls the hardness of cluster assignment; Update cluster center μ k , the calculation formula is:

7. The method for mining regional food culture preferences based on points of interest according to claim 6, characterized in that: The method of continuously updating the embedding representation of nodes by optimizing modularity is: Calculate the gradient propagation and define the function f i,l (μ,x) represents the relationship between the cluster center and the eigenvector, and the calculation formula is: in, is the cluster center μ i The lth component of ; Use the implicit function theorem to calculate the partial derivative of the cluster center with respect to the eigenvector 8. A device for mining regional food culture preferences based on points of interest, characterized in that: include: The first acquisition module is used to preset a number of regions and a number of cuisines, and obtain the preference index of each region for each cuisine; The second acquisition module is used to obtain the amount of population mobility between two regions; Dynamic weighted undirected graph construction module, used to construct a dynamic weighted undirected graph dataset, where regions are used as nodes, preference indexes are used as node features, population migration between regions is used as edges, and the amount of population migration is used as edge weights; Preference identification module, used for graph clustering based on dynamic weighted undirected graph datasets: node features are extracted using a community detection algorithm based on a dynamic graph attention network, several regions are clustered based on the node features, and regions with similar eating habits are identified.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for mining regional food culture preferences based on points of interest as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for mining regional food culture preferences based on points of interest as described in any one of claims 1 to 7 is implemented.

Images