Method and device for recommending points of interest based on point of interest contours

Abstract

The present application relates to a method, apparatus, computer device, and storage medium for recommending points of interest based on point of interest profiles. The method comprises: responding to a reverse geocoding request containing reverse geocoding request coordinates sent by a user terminal, obtaining a plurality of associated points of interest based on the reverse geocoding request coordinates; generating a feature scalar for each associated point of interest based on the reverse geocoding request coordinates and the plurality of associated points of interest; inputting the feature scalar for each associated point of interest into a point of interest recommendation model trained based on point of interest profiles, obtaining recommended points of interest output by the point of interest recommendation model; and sending the recommended points of interest to the user terminal. Embodiments of the present application can improve the accuracy of point of interest recommendations.

Description

Technical Field

[0001] The present application relates to the field of address recommendation, and in particular to a method, apparatus, computer device, and storage medium for recommending points of interest based on point profiles. Background Art

[0002] In today's internet age, with the rapid development of digital maps, people inevitably use the dragging pin feature when using maps. This technology involves reverse geocoding. Reverse geocoding, also known as reverse address lookup or reverse address encoding, essentially resolves the longitude and latitude of a user's pin and returns a corresponding point of interest to the user.

[0003] Currently, when performing reverse geocoding, recommendations are mainly based on information such as parent-child point relationship / spatial index / shortest distance. However, the recommendation accuracy of the above methods is not high. Summary of the Invention

[0004] In response to the above-mentioned deficiencies or shortcomings, the present application provides a method, apparatus, computer device and storage medium for recommending points of interest based on point of interest profiles. The embodiments of the present application can improve the accuracy of point of interest recommendations.

[0005] According to a first aspect, the present application provides a method for recommending points of interest based on point of interest contours. In one embodiment, the method includes: responding to a reverse address coding request containing reverse address coding request coordinates sent by a user terminal, obtaining multiple associated points of interest according to the reverse address coding request coordinates; generating a feature scalar of each associated point of interest according to the reverse address coding request coordinates and the multiple associated points of interest; inputting the feature scalar of each associated point of interest into a point of interest recommendation model trained based on point of interest contours, obtaining recommended points of interest output by the point of interest recommendation model; and sending the recommended points of interest to the user terminal.

[0006] In one embodiment, multiple associated points of interest are obtained based on the reverse address coding request coordinates, including: determining the target geographical range based on the reverse address coding request coordinates; obtaining points of interest whose geographical locations are within the target geographical range from a point of interest library, and using each obtained point of interest as an associated point of interest.

[0007] In one embodiment, the training process of the point of interest recommendation model includes: obtaining request logs of multiple historical requests; the request logs of the historical requests include historical reverse geocoding request coordinates; performing feature engineering based on the request logs of the multiple historical requests to obtain a spatiotemporal feature set; generating training sample data based on the spatiotemporal feature set and all historical reverse geocoding request coordinates; and using the training sample data to train the point of interest recommendation model to obtain a trained point of interest recommendation model.

[0008] In one embodiment, feature engineering is performed based on the request logs of the multiple historical requests, including: preprocessing the request logs of the multiple historical requests, the preprocessing including null value processing, abnormal value processing and / or missing field processing; and feature engineering is performed based on the preprocessed request logs.

[0009] In one embodiment, the POI recommendation model is an ensemble tree model.

[0010] In one embodiment, the spatiotemporal feature set includes multiple spatiotemporal features, including the area of ​​the point of interest contour, the distance between the historical reverse geocoding request coordinates and the point of interest, and the closest distance between the historical reverse geocoding request coordinates and the point of interest contour.

[0011] In one embodiment, the feature scalars of each associated interest point are input into an interest point recommendation model trained based on interest point profiles to obtain recommended interest points output by the interest point recommendation model, including: inputting the feature scalars of each associated interest point into the trained interest point recommendation model; using the trained interest point recommendation model to calculate the confidence corresponding to each input feature scalar, and determining the associated interest point corresponding to the feature scalar with the highest confidence as the recommended interest point output.

[0012] According to a second aspect, the present application provides a device for recommending points of interest based on point of interest profiles. In one embodiment, the device includes:

[0013] an associated interest point acquisition module, configured to respond to a reverse geocoding request including reverse geocoding request coordinates sent by a user terminal, and acquire a plurality of associated interest points according to the reverse geocoding request coordinates;

[0014] A feature scalar generating module, configured to generate a feature scalar of each associated interest point according to the reverse geocoding request coordinates and the plurality of associated interest points;

[0015] A recommendation module, configured to input the characteristic scalars of each associated interest point into an interest point recommendation model trained based on interest point contours, and obtain recommended interest points output by the interest point recommendation model;

[0016] The sending module is used to send the recommended points of interest to the user end.

[0017] According to a third aspect, the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the above-mentioned method embodiments when executing the computer program.

[0018] According to a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any of the above-mentioned method embodiments when the computer program is executed by a processor.

[0019] In an embodiment of the present application, the server responds to a reverse address coding request including reverse address coding request coordinates sent by a user terminal, obtains a plurality of associated points of interest based on the reverse address coding request coordinates; generates a feature scalar of each associated point of interest based on the reverse address coding request coordinates and the plurality of associated points of interest; inputs the feature scalar of each associated point of interest into a point of interest recommendation model trained based on the point of interest contour, obtains a recommended point of interest output by the point of interest recommendation model; and sends the recommended point of interest to the user terminal. When performing reverse address coding, the embodiment of the present application also considers the influence of the point of interest contour area on the point of interest, thereby improving the accuracy of the point of interest recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 1 is a flow chart of a method for recommending points of interest based on point of interest profiles in one embodiment;

[0021] Figure 2 A schematic diagram of an architecture for recommending points of interest in one embodiment;

[0022] Figure 3 Schematic diagram of the model principle in one embodiment;

[0023] Figure 4 is a structural block diagram of a device for recommending points of interest based on point of interest profiles in one embodiment;

[0024] Figure 5 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0026] In the existing technology, reverse geocoding is mainly performed through the following schemes:

[0027] 1. Based on parent-child point relationships: Build parent-child point relationships such as provinces, cities, districts, streets, and roads in advance. When performing reverse geocoding, search for provinces, cities, districts, streets, and roads in sequence based on the coordinates, and concatenate the query results to complete the complete description and return it to the user.

[0028] 2. Spatial indexing: Divide the geographic space into grids in advance, construct an index of grids and coordinates, and create a corresponding address description for each grid. This address description is a virtual point of interest. When performing reverse geocoding, the corresponding grid is found based on the requested latitude and longitude, and the virtual point of interest corresponding to the grid is returned to the user as the reverse geocoding result.

[0029] 3. Based on the shortest distance: Divide the geographic space into grids in advance and build an index of grids and coordinates. When performing reverse geocoding, find the corresponding grid based on the requested longitude and latitude, and then return the point of interest within the grid that is closest to the requested longitude and latitude to the user.

[0030] The above solution 1 can achieve fast search based on the parent-child point relationship established in advance. However, the points of interest returned to the user by solution 1 are gate address points of interest, and the user's understanding of gate addresses is usually vague.

[0031] The above solution 2 can achieve fast search based on the pre-built index and has a fast response speed. However, solution 2 returns virtual points of interest to the user, and some users have cognitive problems that virtual points of interest cannot be matched with actual points of interest.

[0032] The above solution 3 can realize fast search based on the pre-built index. However, when the solution 3 is implemented, it is easy to recommend unpopular points of interest to users, while most users are not familiar with these unpopular points of interest.

[0033] In general, the above-mentioned existing reverse geocoding solutions cannot accurately recommend points of interest to users, that is, the recommended points of interest are not the points of interest that users want (which can be called target points of interest), or are significantly different from the user's target points of interest.

[0034] As previously mentioned, existing reverse geocoding schemes lack high accuracy. Specifically, the inventors recognized that traditional reverse geocoding methods only consider geospatial factors in their design philosophy. This can result in recommendations for low-visibility or virtual POIs. Visibility refers to the user's awareness of a POI; generally, POIs with high awareness also have high visibility.

[0035] For example, a user wants to take a taxi to go shopping, but is unfamiliar with their destination. This user submits a reverse geocoding request to the server. Assume there are two points of interest (POIs) near the user's destination: "Youth Park" and "Lotus Convenience Store." "Youth Park" is a planar area on the map, similar to a regional landmark POI. It's generally well-known to residents of the surrounding area, and even to residents of the city itself. "Lotus Convenience Store" is just a dot on the map, and may be unknown to residents of the surrounding area. For POIs returned by reverse geocoding, users tend to choose POIs with high visibility. These POIs are more likely to be recognized by users, while virtual POIs or POIs with low visibility are less likely to be recognized by users. This may cause users to request reverse geocoding multiple times until they find a POI with high visibility, which can lead to a poor user experience.

[0036] In order to solve the above-mentioned shortcomings, the present application provides a method for recommending points of interest based on the outline of points of interest. In one embodiment, the method for recommending points of interest based on the outline of points of interest includes the following steps: Figure 1 The steps shown are described below using the method applied to a server as an example.

[0037] S110: In response to a reverse geocoding request including reverse geocoding request coordinates sent by a user terminal, a plurality of associated points of interest are acquired according to the reverse geocoding request coordinates.

[0038] The reverse geocoding request may come from a user end, which may be a computing device such as a smartphone, tablet computer, or laptop computer. The requested coordinates may be longitude and latitude information.

[0039] In one embodiment, the above-mentioned acquisition of multiple associated points of interest based on the reverse address coding request coordinates includes: determining the target geographical range based on the reverse address coding request coordinates; acquiring points of interest whose geographical locations are within the target geographical range from the point of interest library, and using each acquired point of interest as an associated point of interest.

[0040] S120: Generate a feature scalar of each associated interest point according to the reverse geocoding request coordinates and the multiple associated interest points.

[0041] S130: Inputting the feature scalar of each associated interest point into an interest point recommendation model trained based on interest point profiles to obtain recommended interest points output by the interest point recommendation model.

[0042] In one embodiment, the aforementioned step of inputting the feature scalars of each associated point of interest into a point of interest recommendation model trained based on point of interest profiles to obtain recommended points of interest output by the point of interest recommendation model includes: inputting the feature scalars of each associated point of interest into the trained point of interest recommendation model; using the trained point of interest recommendation model to calculate the confidence level corresponding to each input feature scalar, and determining the associated point of interest corresponding to the feature scalar with the highest confidence level as the recommended point of interest output. The confidence level refers to the score assigned to the point of interest by the point of interest recommendation model, which indicates the likelihood that the point of interest will satisfy the user, and is typically a floating point number between 0 and 1.

[0043] The visibility of a point of interest (POI) is influenced by many factors, the most important of which is its outline. The inventors have discovered that the larger the outline of a POI, the greater the probability of its recognition by users. Therefore, when training the POI recommendation model, this embodiment not only considers geographic location, such as the distance between the POI and the reverse geocoding request coordinates, but also the area of ​​the POI outline and the closest distance between the reverse geocoding request coordinates and the POI outline. By comprehensively considering these factors, the reverse geocoding effect can be improved, thereby enhancing recommendation accuracy.

[0044] The point of interest recommendation model can be implemented using an integrated tree model. Among them, since this solution integrates factors such as the distance between the point of interest and the reverse address coding request coordinates, the area of ​​the point of interest outline, and the closest distance between the reverse address coding request coordinates and the point of interest outline, it is equivalent to introducing multiple features. In order to be able to comprehensively refer to all features and output a meaningful result, the inventor found that the integrated tree model (also called a tree model) is the best after a large number of experiments. By packaging all features and passing them into the tree model, the tree model makes full use of the structural advantages of the binary tree and the interpretability of the tree model, comprehensively evaluates different features and outputs confidence, and finally sorts the confidence. Through this method, multiple features can be applied to address recommendation scenarios. Finally, through improvements in ideas and implementation methods, the user perception problem of virtual points of interest or unpopular points of interest recommended by traditional methods can be solved.

[0045] S140: Send the recommended points of interest to the user terminal.

[0046] After obtaining the recommended points of interest, the information of the recommended points of interest is returned to the user terminal that sent the reverse geocoding request.

[0047] like Figure 2 As shown, in this embodiment, after receiving the reverse address coding request carrying the reverse address coding request coordinates, the server obtains multiple related interest points according to the reverse address coding request coordinates, generates feature scalars for each related interest point according to the reverse address coding request coordinates and the multiple related interest points, and then inputs the feature scalars for each related interest point into a pre-trained interest point recommendation model, and obtains the recommended interest points output by the interest point recommendation model. Finally, the reverse address coding request can be responded to according to the recommended interest points. When performing reverse address coding, the embodiment of the present application takes into account the influence of the contour area of ​​the interest point on the accuracy of the interest point recommendation, thereby ultimately improving the accuracy of the interest point recommendation, thereby solving the problem of users having cognitive impairment of the interest points returned by the reverse address coding, and fundamentally improving the rationality and usability of the reverse address coding.

[0048] In one embodiment, the training process of the point of interest recommendation model includes: obtaining request logs of multiple historical requests; the request logs of the historical requests include historical reverse geocoding request coordinates; performing feature engineering based on the request logs of the multiple historical requests to obtain a spatiotemporal feature set; generating training sample data based on the spatiotemporal feature set and all historical reverse geocoding request coordinates; and using the training sample data to train the point of interest recommendation model to obtain a trained point of interest recommendation model.

[0049] Among them, the above-mentioned spatiotemporal feature set includes spatiotemporal features corresponding to multiple points of interest, and any spatiotemporal feature includes the contour area of ​​the point of interest, the distance between the historical reverse address coding request coordinates and the point of interest, and the closest distance between the historical reverse address coding request coordinates and the contour of the point of interest.

[0050] Specifically, the above-mentioned root performs feature engineering based on the request logs of the multiple historical requests, including: preprocessing the request logs of the multiple historical requests, the preprocessing including null value processing, abnormal value processing and / or missing field processing; and performing feature engineering based on the preprocessed request logs.

[0051] In order to better understand the above embodiments, the above embodiments are described below through specific application embodiments (which may be referred to as application examples).

[0052] This application example includes four parts: log extraction, feature engineering, model training, and prediction. Each part is introduced below.

[0053] 1. Log extraction

[0054] (1) Log collection

[0055] This application example performs model training based on the historical reverse address encoding request logs of each user. First, it is necessary to collect the historical reverse address encoding logs of each user. Since the amount of data in the collected request logs is very large, the collected request logs can be stored in a distributed file system, and then the request logs in the distributed file system are preprocessed. The above-mentioned preprocessing operations can at least include null value processing, outlier processing and / or missing field processing. The preprocessed request logs can be used to generate model-dependent features.

[0056] (2) Label generation

[0057] This application uses a supervised learning approach, requiring each POI to be labeled and categorized as either positive or negative. Positive examples represent reverse-geocoded POIs that the user accepts, while negative examples represent reverse-geocoded POIs that the user dislikes. Throughout the training process, labels are used as a benchmark to determine the category of each POI.

[0058] 2. Feature Engineering

[0059] Historical reverse geocoding request logs contain a wealth of user behaviors and operating habits, some of which are difficult to detect but implicit in the data. By breaking down the data into different dimensions and extracting features, we can generate features of various dimensions, such as distance and the area of ​​the POI outline (which can be calculated using geometric formulas). Some of these features are empirically helpful in determining which POI a user prefers, and some exhibit a certain correlation with the user's final choice in terms of correlation coefficients (such as covariance and Pierce coefficient).

[0060] Specifically, spatiotemporal features, such as distance and user historical behavior popularity, are extracted from preprocessed request logs. Furthermore, multi-time dimension log aggregation projects can be further utilized to aggregate request logs of time dimensions such as 30 days and 60 days, thereby obtaining features such as user historical behavior popularity over 30 days and 60 days.

[0061] Furthermore, taking the historical request coordinates in each historical request (referring to the reverse address coding request in the completed order) as the unit, each historical reverse address coding request coordinate and all points of interest within a diameter of 1000 meters (just as an example) of each historical reverse address coding request coordinate (these points of interest can be called associated points of interest, for example, taking the historical reverse address coding request coordinate A as an example, each point of interest whose geographical location is within an area with a diameter of 1000 meters with the historical reverse address coding request coordinate A as the center is an associated point of interest of the historical reverse address coding request coordinate A) as the basic data for generating features, we can eventually generate a feature scalar for each historical reverse address coding request coordinate for each of its associated points of interest, that is, <reverse address coding request coordinate, <point of interest id, feature>>.

[0062] A training sample can be extracted from the request log of each historical request. After the training samples are extracted from the request logs of multiple historical reverse geocoding requests, the next step of model training can be performed based on the extracted training samples.

[0063] 3. Model Training

[0064] Using the features contained in each training sample as input and the mean squared error (MSE) as the objective function, the algorithm seeks the parameters that minimize the objective function. When the objective function reaches its minimum, a trained POI recommendation model is obtained based on the parameters at that point. This model can be used to predict the category of unknown inputs and outputs a floating-point "confidence score" between 0 and 1. The POI with the highest "confidence score" is returned to the user.

[0065] Specifically, this application example uses the "ensemble tree model" as the point of interest recommendation model. The model training process can be found in Figure 3The "ensemble tree model" is composed of multiple "binary tree" data structures (i.e., a root node and at most two child nodes, called left nodes and right nodes, respectively) connected in sequence. Each "root node" in a formal sense is used to store the division threshold (i.e., condition), and the two child nodes are used to store data that meets the condition and data that does not meet the condition, respectively. For the t-th tree, the information received is the numerical error between the output result of the t-1-th tree for a certain feature (a floating point number between 0 and 1, indicating the possibility) and the label (0 or 1). The ultimate goal of the model is to reduce the overall error and make the value output by the model closer to the true label, so as to have a highly reliable classification ability for unknown data.

[0066] 4. Prediction

[0067] Prediction is the process of classification. For a piece of real data (<user uid, reverse geocoding request coordinates, points of interest within a 1000m diameter>), the same feature engineering method is used to convert the reverse geocoding request coordinates and all points of interest within a 1000m diameter of the reverse geocoding request coordinates into feature scalars. The generated feature scalar pairs are input into the trained model. The model will derive the confidence of each point of interest based on the input feature scalars and output the point of interest with the highest confidence.

[0068] Figure 1 FIG. 1 is a flow chart of a method for recommending points of interest based on the outline of points of interest in one embodiment. Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0069] Based on the same inventive concept, the present application also provides a device for recommending points of interest based on the outline of the points of interest. Figure 4 As shown, the device for recommending points of interest based on point of interest profiles includes the following modules:

[0070] The associated interest point acquisition module 110 is configured to respond to a reverse geocoding request including reverse geocoding request coordinates sent by a user terminal, and acquire a plurality of associated interest points according to the reverse geocoding request coordinates;

[0071] A feature scalar generating module 120 is configured to generate a feature scalar of each associated interest point according to the inverse geocoding request coordinates and the plurality of associated interest points;

[0072] The recommendation module 130 is configured to input the characteristic scalar of each associated interest point into an interest point recommendation model trained based on interest point profiles, and obtain recommended interest points output by the interest point recommendation model;

[0073] The sending module 140 is configured to send the recommended points of interest to the user terminal.

[0074] In one embodiment, the associated interest point acquisition module 110 is used to determine the target geographical range according to the reverse geocoding request coordinates; obtain interest points whose geographical locations are within the target geographical range from the interest point library, and use each obtained interest point as an associated interest point.

[0075] In one embodiment, the above device further includes a training module.

[0076] A training module is used to obtain request logs of multiple historical requests; the request logs of historical requests include historical reverse geocoding request coordinates; feature engineering is performed based on the request logs of the multiple historical requests to obtain a spatiotemporal feature set; training sample data is generated based on the spatiotemporal feature set and all historical reverse geocoding request coordinates; and a point of interest recommendation model is trained using the training sample data to obtain a trained point of interest recommendation model.

[0077] In one embodiment, when the training module performs feature engineering based on the request logs of the multiple historical requests, it is specifically used to preprocess the request logs of the multiple historical requests, and the preprocessing includes null value processing, outlier processing and / or missing field processing; feature engineering is performed based on the preprocessed request logs.

[0078] In one embodiment, the POI recommendation model is an ensemble tree model.

[0079] In one embodiment, the spatiotemporal feature set includes multiple spatiotemporal features, including the area of ​​the point of interest contour, the distance between the historical reverse geocoding request coordinates and the point of interest, and the closest distance between the historical reverse geocoding request coordinates and the point of interest contour.

[0080] In one embodiment, the recommendation module 130 is used to input the feature scalars of each associated interest point into a trained interest point recommendation model; use the trained interest point recommendation model to calculate the confidence corresponding to each input feature scalar, and determine the associated interest point corresponding to the feature scalar with the highest confidence as the recommended interest point output. For the specific definition of the interest point recommendation device based on the interest point profile, please refer to the definition of the interest point recommendation method based on the interest point profile above, which will not be repeated here. The various modules in the above-mentioned interest point recommendation device based on the interest point profile can be fully or partially implemented by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

[0081] In one embodiment, a computer device is provided, whose internal structure diagram can be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as request logs. The specific stored data can also be found in the definitions in the above method embodiments. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for recommending points of interest based on the profile of the points of interest is implemented.

[0082] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0083] This embodiment also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the method provided in any of the above method embodiments are implemented.

[0084] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the method provided in any of the above method embodiments are implemented.

[0085] Those skilled in the art will appreciate that all or part of the processes in the above method embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0086] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0087] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for recommending points of interest based on point of interest contours, characterized in that: The method comprises: In response to a reverse geocoding request including reverse geocoding request coordinates sent by a user terminal, obtaining a plurality of associated points of interest according to the reverse geocoding request coordinates; generating a feature scalar of each associated interest point according to the reverse geocoding request coordinates and the plurality of associated interest points; Inputting the characteristic scalar of each of the associated interest points into an interest point recommendation model trained based on interest point profiles to obtain recommended interest points output by the interest point recommendation model; Sending the recommended points of interest to the user terminal; The training process of the POI recommendation model includes: Obtaining a request log of a plurality of historical requests; the request log of the historical requests including historical reverse geocoding request coordinates; Perform feature engineering based on the request logs of the multiple historical requests to obtain a spatiotemporal feature set; generating training sample data based on the spatiotemporal feature set and all historical reverse geocoding request coordinates; Using the training sample data to train a point of interest recommendation model to obtain a trained point of interest recommendation model; The spatiotemporal feature set includes multiple spatiotemporal features, and the spatiotemporal features include the area of ​​the interest point outline, the distance between the historical reverse geocoding request coordinates and the interest point, and the closest distance between the historical reverse geocoding request coordinates and the interest point outline.

2. The method according to claim 1, wherein The acquiring of a plurality of associated points of interest according to the reverse geocoding request coordinates includes: Determine the target geographical area according to the reverse geocoding request coordinates; The points of interest whose geographical locations are within the target area are acquired from the point of interest database, and each acquired point of interest is used as an associated point of interest.

3. The method according to claim 1, wherein The performing feature engineering according to the request logs of the multiple historical requests includes: Preprocessing the request logs of the multiple historical requests, wherein the preprocessing includes null value processing, abnormal value processing, and / or missing field processing; Perform feature engineering based on preprocessed request logs.

4. The method according to claim 1, wherein The point of interest recommendation model is an integrated tree model.

5. The method according to claim 1, wherein Inputting the characteristic scalar of each associated interest point into an interest point recommendation model trained based on interest point profiles to obtain recommended interest points output by the interest point recommendation model includes: Inputting the characteristic scalar of each associated interest point into the trained interest point recommendation model; The trained POI recommendation model is used to calculate the confidence corresponding to each input feature scalar, and the associated POI corresponding to the feature scalar with the highest confidence is determined as the recommended POI output.

6. A device for recommending points of interest based on point of interest contours, characterized in that: The device comprises: an associated interest point acquisition module, configured to respond to a reverse geocoding request including reverse geocoding request coordinates sent by a user terminal, and acquire a plurality of associated interest points according to the reverse geocoding request coordinates; a feature scalar generating module, configured to generate a feature scalar of each of the associated interest points according to the reverse address coding request coordinates and the multiple associated interest points; A recommendation module, configured to input the characteristic scalar of each of the associated interest points into an interest point recommendation model trained based on interest point profiles, and obtain recommended interest points output by the interest point recommendation model; A sending module, configured to send the recommended points of interest to the user terminal; The training process of the POI recommendation model includes: Obtaining a request log of a plurality of historical requests; the request log of the historical requests including historical reverse geocoding request coordinates; Perform feature engineering based on the request logs of the multiple historical requests to obtain a spatiotemporal feature set; generating training sample data based on the spatiotemporal feature set and all historical reverse geocoding request coordinates; Using the training sample data to train a point of interest recommendation model to obtain a trained point of interest recommendation model; The spatiotemporal feature set includes multiple spatiotemporal features, and the spatiotemporal features include the area of ​​the interest point outline, the distance between the historical reverse geocoding request coordinates and the interest point, and the closest distance between the historical reverse geocoding request coordinates and the interest point outline.

7. A computer 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 computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images