Method, device, system and medium for recommending points of interest based on user check-in data

By constructing user-interest point graphs and time-interest point graphs, combining graph convolution neural network model and geographical factors, the user's preference score for interest points is calculated, and the problems of data sparseness and complex context in point recommendations are solved. , achieving accurate user preference capture and recommendation performance improvements.

CN115033810BActive Publication Date: 2024-12-03WUHAN UNIV OF TECH
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202210697215.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-06-20
Publication Date
2024-12-03
Estimated Expiration
2042-06-20

AI Technical Summary

Technical Problem

In location-based social networks, point-of-interest recommendations face problems of data sparsity and complex contexts, and it is difficult to accurately capture user's point-of-interest preferences.

Method used

By constructing user-interest point graphs and time-interest point graphs, extracting the feature vectors of users and interest points, calculating geographic factor coefficients, and using the improved graph convolutional neural network model combined with a unified scoring formula, calculate user preferences for interest points Score, generate a recommended sequence.

Benefits of technology

Accurately capture user preferences in complex context environments, improve the performance of point-of-interest recommendations, and effectively solve the problems of data sparseness and the impact of multiple contexts.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115033810B_ABST
    Figure CN115033810B_ABST
Patent Text Reader

Abstract

The present application provides a method for recommending points of interest based on user check-in data, including: obtaining a user check-in set, and constructing a user-point of interest map and a time-point of interest map according to the user check-in set; extracting feature vectors of users and points of interest from the user-point of interest map, and extracting feature vectors of time and points of interest from the time-point of interest map; calculating the geographic factor coefficient of the points of interest; calculating the preference score of the target user for each point of interest; and generating a recommendation sequence for the target according to the preference score of the target user for each point of interest. The present invention comprehensively considers the influence of time factors and geographic factors, and introduces them into a unified model to jointly model user preferences, improves the recommendation performance of the model, and can accurately capture user preferences in complex context environments, and has good practical value.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of point of interest recommendation, and in particular to a method, device, system and medium for recommending points of interest based on user check-in data. Background Art

[0002] In location-based social networks, users communicate with other users on the platform by sharing check-in information. The core content of check-in information is the check-in location, which is called a point of interest (POI) in location-based social networks. With the rapid development of location-based social networks, the scale of social platforms continues to expand. On the one hand, large-scale platforms have attracted many businesses to settle in, which has increased the types and range of choices for users, leading to the problem of information overload - users find it difficult to make efficient choices in a short time; on the other hand, large-scale platforms can carry more active users, and these active users will generate massive amounts of user behavior data during their interaction with social platforms. Obviously, fully mining user behavior data can create huge economic value for the platform. In order to improve the user experience and increase the economic benefits of the platform, point of interest recommendation came into being and gradually became a hot topic of research.

[0003] Compared with traditional recommendation tasks, such as music recommendation and movie recommendation, the challenges faced by POI recommendation are more severe. The challenges mainly include two points: first, data sparsity. In the scenario of POI recommendation, the cost of generating an activity record for a user is high, which leads to fewer check-in records for most users, that is, the modeling will face a more serious data sparsity problem; second, complex context. When a user makes a check-in decision, it will not only be affected by the time context, but also by the geographical context. This also means that the user's decision will be affected by multiple contexts, that is, the user's preference will change with the change of the context. How to accurately capture the user's preference in a complex context is another major challenge in the field of POI recommendation. Summary of the invention

[0004] The purpose of the present invention is to overcome the above technical deficiencies, provide a method for recommending points of interest based on user check-in data, and solve the technical problem in the prior art that it is difficult to accurately capture the user's point of interest preferences in a complex context environment.

[0005] In order to achieve the above technical objectives, in a first aspect, the technical solution of the present invention provides a method for recommending points of interest based on user check-in data, comprising the following steps:

[0006] Obtain a user check-in set, and construct a user-interest point graph and a time-interest point graph according to the user check-in set;

[0007] Extracting feature vectors of users and points of interest from the user-point of interest graph, wherein the feature vectors of users and points of interest are used to describe the association relationship between users and points of interest, and extracting feature vectors of time and points of interest from the time-point of interest graph, wherein the feature vectors of time and points of interest are used to describe the association relationship between time and points of interest;

[0008] Calculate the geographical factor coefficient of the point of interest;

[0009] Calculate the target user's preference score for each point of interest based on the feature vectors of the user and the point of interest, the feature vectors of the time and the point of interest, and the geographic factor coefficient of the point of interest;

[0010] A recommendation sequence for the target is generated according to the target user's preference score for each point of interest.

[0011] Compared with the prior art, the beneficial effects of the present invention include:

[0012] The method for recommending points of interest based on user check-in data provided by the present invention constructs two bipartite graphs according to the user's check-in to capture the correlation between points of interest and users and between points of interest and time factors respectively; then, the improved graph convolutional neural network model is used to extract users, points of interest and multiple context features respectively; finally, a unified scoring formula is used to calculate the preference score for each point of interest for the target user, and a recommendation list is generated. In the field of point of interest recommendation, time factors and geographical factors play an important role in modeling user preferences. The present invention comprehensively considers the influence of the above two factors and introduces them into a unified model to jointly model user preferences, thereby improving the recommendation performance of the model, being able to accurately capture user preferences in complex context environments, and having good practical value.

[0013] According to some embodiments of the present invention, generating a recommendation sequence for the target user according to the target user's preference score for each point of interest comprises the steps of:

[0014] The interest points are sorted in descending order according to the preference scores, and the first k interest points are taken to form a recommendation sequence for the target user.

[0015] According to some embodiments of the present invention, after calculating the target user's preference score for each point of interest based on the feature vectors of the user and the point of interest, the feature vectors of the time and the point of interest, and the geographical factor coefficient of the point of interest, the steps include:

[0016] The model parameters are optimized using the stochastic gradient descent algorithm.

[0017] According to some embodiments of the present invention, the step of calculating the geographical factor coefficient of the point of interest comprises the following steps:

[0018] Get geographic distance function and geographic distance threshold;

[0019] Calculate a geographic distance coefficient according to the geographic distance function and the geographic distance threshold;

[0020] The geographic factor coefficient of the point of interest is calculated according to the geographic distance coefficient.

[0021] According to some embodiments of the present invention, the geographic distance coefficient calculated according to the geographic distance function and the geographic distance threshold is expressed as:

[0022]

[0023] dis(l,l p ) represents the geographic distance coefficient between the point of interest p and the current location l, geo(l,l p ) represents the geographical distance function between two points calculated by longitude and latitude, d o Indicates the geographic distance threshold, max(.) means taking the maximum value;

[0024] The geographic factor coefficient of the point of interest calculated according to the geographic distance coefficient is expressed as:

[0025]

[0026] According to some embodiments of the present invention, the preference score of the target user for each point of interest is calculated based on the feature vectors of the user and the point of interest, the feature vectors of the time and the point of interest, and the geographical factor coefficient of the point of interest, and is expressed as follows:

[0027]

[0028] Among them, time is t x , position is l o , user is u i , the point of interest is p j .

[0029] According to some embodiments of the present invention, the model parameters are optimized using a stochastic gradient descent algorithm, comprising the steps of:

[0030] Optimize model parameters using a BPR-based loss function:

[0031]

[0032] Among them, D u represents the check-in set of user u, τ(.) represents the sigmoid function, λ represents the regularization coefficient, Θ represents the model parameter, is the preference score of the interest point pj, Point of interest p i preference score.

[0033] In a second aspect, the present invention provides a device for recommending points of interest based on user check-in data, comprising:

[0034] An interest point graph construction module is used to obtain a user check-in set and construct a user-interest point graph and a time-interest point graph according to the user check-in set;

[0035] A graph structure feature extraction module, configured to extract feature vectors of users and points of interest from the user-point of interest graph, wherein the feature vectors of users and points of interest are used to describe the association relationship between users and points of interest, and to extract feature vectors of time and points of interest from the time-point of interest graph, wherein the feature vectors of time and points of interest are used to describe the association relationship between time and points of interest;

[0036] A geographical factor evaluation module is used to calculate the geographical factor coefficient of the point of interest;

[0037] A score calculation module, used to calculate the target user's preference score for each point of interest based on the feature vectors of the user and the point of interest, the feature vectors of the time and the point of interest, and the geographical factor coefficient of the point of interest;

[0038] The recommendation module is used to generate a recommendation sequence for the target according to the preference score of the target user for each point of interest.

[0039] In a third aspect, the present invention provides a point of interest recommendation system based on user check-in data, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the method for recommending points of interest based on user check-in data as described in any one of the first aspects is implemented.

[0040] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions for causing a computer to execute a method for recommending points of interest based on user check-in data as described in any one of the first aspects.

[0041] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, wherein the abstract drawings are identical to one of the drawings in the specification:

[0043] Figure 1 A flowchart of a method for recommending points of interest based on user check-in data provided by an embodiment of the present invention;

[0044] Figure 2 A flowchart of a method for recommending points of interest based on user check-in data provided by another embodiment of the present invention;

[0045] Figure 3 A flowchart of a method for recommending points of interest based on user check-in data provided by another embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0047] It should be noted that, although the functional modules are divided in the system schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the system or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0048] The present invention provides a method for recommending points of interest based on user check-in data, which comprehensively considers the influence of time factors and geographical factors, and introduces them into a unified model to jointly model user preferences, thereby improving the recommendation performance of the model. It can accurately capture user preferences in complex contextual environments and has good practical value.

[0049] In the direction of POI research, research work in recent years has deeply explored the impact of multiple factors on user behavior, such as time factors, geographical factors and POI classification. In the related technology, the user's check-in time is divided into two models, weekdays and weekends, to explore the time correlation of user check-in behavior, and the kernel density estimation method is used to model the impact of geographical factors on user preferences. Finally, based on the above two, a recommendation model TICRec that integrates time correlation and regional correlation is proposed. In the related technology, the impact of time factors on user decision-making is deeply explored, and a double-weighted low-rank graph construction model is proposed. This model combines user interests and their changing order preferences with time interval evaluation to provide POI recommendations for specific time periods. In the related technology, the location of user check-in behavior is visualized and analyzed, and a spatial aggregation phenomenon is found, that is, individual visit locations tend to cluster together, and user activity areas and POI influence areas are defined to capture this phenomenon. In the related technology, the idea of ​​ranking learning is introduced into the field of POI recommendation, and the user's check-in frequency is used to construct a partial order relationship for POIs, and a ranking-based geographic factor decomposition model Rank-GeoFM is proposed to alleviate the negative impact of data sparsity on the recommendation performance of the model. In the related art, the representation learning-based method proposes a recommendation model that jointly models the influence of user preferences and the sequential transformation of points of interest by incorporating the influence of geographical factors. In the related art, time factors and space factors are integrated to propose a model UPOST that establishes user preferences for spatiotemporal themes. This model infers the user's preferences for different types of places at different times by learning spatiotemporal themes from the user's historical semantic positions.

[0050] From the above research, we can see that in the field of POI recommendation, time factors and geographical factors play an important role in modeling user preferences. Therefore, this model will comprehensively consider the influence of the above two factors and introduce them into a unified model to jointly model user preferences to improve the recommendation performance of the model.

[0051] The Graph Convolutional Network (GCN) was first proposed by Kipf et al. and used for semi-supervised learning tasks. It aggregates features from the neighboring nodes of the target node through a message passing mechanism to enhance the feature representation of the target node. Although GCN has a powerful ability to extract node features from the graph structure, it is still subject to the influence of the graph scale and aggregation method. In order to solve the problem that GCN cannot be applied to large-scale graph data, the GraphSAGE model is proposed in the related art. The aggregation method in the traditional GCN is changed to sampling aggregation. According to the topological structure of the target node in the network, a certain number of nodes are randomly sampled from its two-step neighboring nodes for feature aggregation, thereby reducing the computational scale of the aggregation operation. Starting from the perspective of the aggregation method, the related art introduces the attention mechanism and constructs the Graph Attention Network (GAT). Different from the average aggregation method in the traditional GCN, GAT learns the attention coefficient between nodes through a layer of feedforward neural network, and uses the coefficient in the aggregation operation to realize weighted feature aggregation. In addition, the related technology starts from the efficiency problem of GCN, points out that there are redundant feature mapping operations in traditional GCN, and proposes a simplified version of GCN, called LightGCN. In LightGCN, only the node aggregation operation is retained. Experimental results show that LightGCN has higher training efficiency without losing model accuracy.

[0052] GCN was inspired by Convolutional Neural Networks (CNN). The core modules of CNN are convolutional layer and pooling layer. Convolutional layer is used for feature extraction, and pooling layer is used for feature compression. The combination of these two can show powerful feature extraction capabilities on data based on Euclidean space, but cannot work on non-Euclidean space data such as graph data. For this reason, Kipf et al. redefined the core operations of CNN, convolution and pooling, on graph-structured data.

[0053] For a given static undirected graph G = (V, E), where V represents the set of nodes in the graph and E represents the set of edges in the graph. i ∈V is the feature representation vector of the h+1th GCN The calculation method is shown in formula (1):

[0054]

[0055] Where N(v i ) represents the node v iIn the first-order neighbor nodes in the figure, the denominator of the fraction in the summation of formula (1) represents the degree normalization of the two nodes, which corresponds to the convolution operation in CNN, that is, the feature extraction operation. h+1 Represents the weight matrix of GCN at the h+1th level, used for feature mapping of feature vectors. σ(.) represents a nonlinear activation function.

[0056] The related technology analyzes the convolution operation of GCN from the perspective of efficiency, and proposes a lightweight graph convolution operation (Light Graph Convolution, LightGC) through theoretical and experimental verification. The calculation method of this operation is shown in formula (2):

[0057]

[0058] From formula (2), we can see that the feature mapping step in traditional GCN is omitted in LightGC, and only the operation of aggregating features from the node neighbor set is retained. From theoretical analysis and experimental data, we can see that omitting feature mapping does not affect the performance of GCN in recommendation tasks, and because this step is omitted, the model overhead is also reduced.

[0059] The embodiments of the present invention are further described below in conjunction with the accompanying drawings.

[0060] Reference Figure 1 , Figure 1 A flowchart of a method for recommending points of interest based on user check-in data provided by an embodiment of the present invention; the method for recommending points of interest based on user check-in data includes but is not limited to the following steps:

[0061] Step S110, obtaining a user check-in set, and constructing a user-interest point graph and a time-interest point graph according to the user check-in set;

[0062] Step S120, extracting feature vectors of users and points of interest from the user-point of interest graph, and extracting feature vectors of time and points of interest from the time-point of interest graph;

[0063] Step S130, calculating and obtaining the geographical factor coefficient of the point of interest;

[0064] Step S140, calculating the preference score of the target user for each point of interest based on the feature vectors of the user and the point of interest, the time and the feature vector of the point of interest, and the geographical factor coefficient of the point of interest;

[0065] Step S150, generating a recommendation sequence for the target according to the target user's preference score for each point of interest.

[0066] In one embodiment, a method for recommending points of interest based on user check-in data includes the steps of: obtaining a user check-in set, and constructing a user-point of interest graph and a time-point of interest graph based on the user check-in set; extracting feature vectors of users and points of interest from the user-point of interest graph, the feature vectors of users and points of interest being used to describe the association between users and points of interest, extracting feature vectors of time and points of interest from the time-point of interest graph, the feature vectors of time and points of interest being used to describe the association between time and points of interest; calculating geographic factor coefficients of points of interest; calculating a preference score of a target user for each point of interest based on the feature vectors of users and points of interest, the feature vectors of time and points of interest, and the geographic factor coefficients of points of interest; and generating a recommendation sequence for the target based on the preference score of the target user for each point of interest.

[0067] The method for recommending points of interest based on user check-in data provided in this embodiment constructs two bipartite graphs according to the user's check-in to capture the association between points of interest and users and between points of interest and time factors respectively; then, the improved graph convolutional neural network model is used to extract users, points of interest and multiple context features respectively; finally, a unified scoring formula is used to calculate the preference score for each point of interest for the target user and generate a recommendation list. In the field of point of interest recommendation, time factors and geographical factors play an important role in modeling user preferences. The present invention comprehensively considers the influence of the above two factors and introduces them into a unified model to jointly model user preferences, thereby improving the recommendation performance of the model, being able to accurately capture user preferences in complex context environments, and having great practical value.

[0068] The graph structure feature extraction module is used to extract node features from two bipartite graphs. Specifically, the feature vectors of users and interest points are extracted from the user-interest point graph. Used to describe the relationship between users and points of interest; extract feature vectors of time and points of interest from the time-point of interest graph Used to describe the relationship between time and points of interest.

[0069] The LightGC operation can efficiently complete the node feature aggregation and extraction operations, but the important attribute of the edge weight in the graph is not combined in LightGC. Whether it is a user-POI graph or a time-POI graph, the weight of the edge in the graph is an extremely important attribute: for the user-POI graph, the edge weight means the frequency of users visiting the POI. The higher the weight, the higher the frequency, that is, the user likes the POI more; for the time-POI graph, the edge weight means the frequency of the POI being visited during the time. The higher the weight, the higher the frequency, which also indicates that the probability of the POI being visited is high during the same time in the next cycle.

[0070] Therefore, in order to incorporate the important attribute of edge weight into the aggregation operation, formula (2) is improved. On the basis of retaining degree normalization, a convolution operation WD-LightGC that integrates degree and weight normalization is proposed to achieve the goal of integrating edge weights:

[0071]

[0072] Formula (3) is the calculation formula of WD-LightGC. By introducing the edge weight coefficient, the node feature extraction process is optimized, thereby improving the recommendation performance of the model. By using formula (3), the corresponding node feature vectors can be extracted from two bipartite graphs.

[0073] Geographical factor evaluation module: Geographical distance is an important factor that affects user decision-making. From the analysis of real-life scenarios, when users are within an acceptable distance range, the impact of geographical factors may be small. However, beyond a certain range, due to the high cost of reaching a location, even if the user likes it very much, he will not visit it. For example, at lunch time, if the restaurant that the user expects is 20 kilometers away from the current location, the user is unlikely to go to this restaurant.

[0074] In order to capture the impact of geographical factors on user check-in behavior, we incorporate a geographical factor evaluation module into RMGCN to calculate the geographical factor coefficient of the current point of interest p. The specific calculation method is shown in formula (4):

[0075]

[0076] where dis(l,l p ) represents the geographic distance coefficient between the point of interest p and the current location l, and its calculation method is shown in formula (5):

[0077]

[0078] where geo(l,l p ) represents a function that calculates the geographical distance between two points using longitude and latitude, both of which are expressed in longitude and latitude. o represents the geographic distance threshold, which is a hyperparameter of the model, and max(.) represents the maximum value. From formula (4) and formula (5), we can see that in d o Within the range, geographical factors have little influence on the user's choice, and gc is a constant; within d o Outside the range, gc will follow dis(l,l p ) value increases, which means that the probability of users visiting places that are too far away is smaller.

[0079] Scoring calculation module: Based on node features and geographic factor coefficients, a scoring calculation module integrating temporal context and spatial context is designed. Specifically, at time t x and location o When user u i and points of interest j The calculation method of preference score is shown in formula (6):

[0080]

[0081] Through formula (6), we can get the target user’s preference score for each point of interest at the current time and geographical location. Then, we can sort them in descending order according to the preference scores to generate a recommendation list for the user at the current time and location.

[0082] Model optimization and recommendation process: In terms of model optimization, a framework based on Bayesian Personalized Ranking (BPR) is used to learn model parameters.

[0083] The loss function based on BPR is defined as shown in formula (7):

[0084]

[0085] Where D u represents the check-in set of user u, τ(.) represents the sigmoid function, λ represents the regularization coefficient, Θ represents the model parameter, i.e., the initialization feature vector of the node in the graph, The point of interest is p j The preference score of Point of interest p i The preference score of the model can be trained by using the stochastic gradient descent algorithm and minimizing formula (7).

[0086] Reference Figure 2 , Figure 2 A flowchart of a method for recommending points of interest based on user check-in data provided by another embodiment of the present invention; the method for recommending points of interest based on user check-in data includes but is not limited to the following steps:

[0087] Step S210, sorting the interest points in descending order according to the preference scores, and taking the first k interest points to form a recommendation sequence for the target user.

[0088] In one embodiment, a method for recommending points of interest based on user check-in data includes the steps of: obtaining a user check-in set, and constructing a user-point of interest graph and a time-point of interest graph based on the user check-in set; extracting feature vectors of users and points of interest from the user-point of interest graph, the feature vectors of users and points of interest being used to describe the association between users and points of interest, extracting feature vectors of time and points of interest from the time-point of interest graph, the feature vectors of time and points of interest being used to describe the association between time and points of interest; calculating geographic factor coefficients of points of interest; calculating a preference score of a target user for each point of interest based on the feature vectors of users and points of interest, the feature vectors of time and points of interest, and the geographic factor coefficients of points of interest; and generating a recommendation sequence for the target based on the preference score of the target user for each point of interest.

[0089] Generate a recommendation sequence for the target user according to the target user's preference score for each point of interest, including the steps of: sorting the points of interest in descending order according to the preference score, and taking the first k points of interest to form a recommendation sequence for the target user. The value range of the recommendation sequence can be k = (5, 10, 15, 20), of course, this embodiment does not constitute a limitation to it, and can be set to other values ​​according to requirements.

[0090] In one embodiment, a method for recommending points of interest based on user check-in data includes the steps of: obtaining a user check-in set, and constructing a user-point of interest graph and a time-point of interest graph based on the user check-in set; extracting feature vectors of users and points of interest from the user-point of interest graph, the feature vectors of users and points of interest are used to describe the association between users and points of interest, extracting feature vectors of time and points of interest from the time-point of interest graph, the feature vectors of time and points of interest are used to describe the association between time and points of interest; calculating geographic factor coefficients of points of interest; calculating the preference score of the target user for each point of interest based on the feature vectors of users and points of interest, the feature vectors of time and points of interest, and the geographic factor coefficients of points of interest; optimizing model parameters using a stochastic gradient descent algorithm; and generating a recommendation sequence for the target based on the preference score of the target user for each point of interest.

[0091] Reference Figure 3 , Figure 3 A flowchart of a method for recommending points of interest based on user check-in data provided by another embodiment of the present invention; the method for recommending points of interest based on user check-in data includes but is not limited to the following steps:

[0092] Step S310, obtaining a geographic distance function and a geographic distance threshold;

[0093] Step S320, calculating a geographic distance coefficient according to the geographic distance function and the geographic distance threshold;

[0094] Step S330, calculating the geographical factor coefficient of the point of interest according to the geographical distance coefficient.

[0095] In one embodiment, a method for recommending points of interest based on user check-in data includes the steps of: obtaining a user check-in set, and constructing a user-point of interest graph and a time-point of interest graph based on the user check-in set; extracting feature vectors of users and points of interest from the user-point of interest graph, the feature vectors of users and points of interest being used to describe the association between users and points of interest, extracting feature vectors of time and points of interest from the time-point of interest graph, the feature vectors of time and points of interest being used to describe the association between time and points of interest; calculating geographic factor coefficients of points of interest; calculating a preference score of a target user for each point of interest based on the feature vectors of users and points of interest, the feature vectors of time and points of interest, and the geographic factor coefficients of points of interest; and generating a recommendation sequence for the target based on the preference score of the target user for each point of interest.

[0096] Calculating the geographic factor coefficient of the point of interest includes the steps of: obtaining a geographic distance function and a geographic distance threshold; calculating the geographic distance coefficient according to the geographic distance function and the geographic distance threshold; and calculating the geographic factor coefficient of the point of interest according to the geographic distance coefficient.

[0097] In one embodiment, a method for recommending points of interest based on user check-in data includes the steps of: obtaining a user check-in set, and constructing a user-point of interest graph and a time-point of interest graph based on the user check-in set; extracting feature vectors of users and points of interest from the user-point of interest graph, the feature vectors of users and points of interest being used to describe the association between users and points of interest, extracting feature vectors of time and points of interest from the time-point of interest graph, the feature vectors of time and points of interest being used to describe the association between time and points of interest; calculating geographic factor coefficients of points of interest; calculating a preference score of a target user for each point of interest based on the feature vectors of users and points of interest, the feature vectors of time and points of interest, and the geographic factor coefficients of points of interest; and generating a recommendation sequence for the target based on the preference score of the target user for each point of interest.

[0098] Calculating the geographic factor coefficient of the point of interest includes the steps of: obtaining a geographic distance function and a geographic distance threshold; calculating the geographic distance coefficient according to the geographic distance function and the geographic distance threshold; and calculating the geographic factor coefficient of the point of interest according to the geographic distance coefficient.

[0099] The geographic distance coefficient calculated based on the geographic distance function and the geographic distance threshold is expressed as:

[0100]

[0101] dis(l,l p) represents the geographic distance coefficient between the point of interest p and the current location l, geo(l,l p ) represents the geographical distance function between two points calculated by longitude and latitude, d o Indicates the geographic distance threshold, max(.) means taking the maximum value;

[0102] The geographic factor coefficient of the point of interest calculated based on the geographic distance coefficient is expressed as:

[0103]

[0104] In one embodiment, a method for recommending points of interest based on user check-in data includes the steps of: obtaining a user check-in set, and constructing a user-point of interest graph and a time-point of interest graph based on the user check-in set; extracting feature vectors of users and points of interest from the user-point of interest graph, the feature vectors of users and points of interest being used to describe the association between users and points of interest, extracting feature vectors of time and points of interest from the time-point of interest graph, the feature vectors of time and points of interest being used to describe the association between time and points of interest; calculating geographic factor coefficients of points of interest; calculating a preference score of a target user for each point of interest based on the feature vectors of users and points of interest, the feature vectors of time and points of interest, and the geographic factor coefficients of points of interest; and generating a recommendation sequence for the target based on the preference score of the target user for each point of interest.

[0105] The target user's preference score for each point of interest is calculated based on the feature vectors of the user and the point of interest, the time and the feature vector of the point of interest, and the geographic factor coefficient of the point of interest, as shown below:

[0106]

[0107] Among them, time is t x , position is l o , user is u i , the point of interest is p j .

[0108] In one embodiment, a method for recommending points of interest based on user check-in data includes the steps of: obtaining a user check-in set, and constructing a user-point of interest graph and a time-point of interest graph based on the user check-in set; extracting feature vectors of users and points of interest from the user-point of interest graph, the feature vectors of users and points of interest are used to describe the association between users and points of interest, extracting feature vectors of time and points of interest from the time-point of interest graph, the feature vectors of time and points of interest are used to describe the association between time and points of interest; calculating geographic factor coefficients of points of interest; calculating the preference score of the target user for each point of interest based on the feature vectors of users and points of interest, the feature vectors of time and points of interest, and the geographic factor coefficients of points of interest; optimizing model parameters using a stochastic gradient descent algorithm; and generating a recommendation sequence for the target based on the preference score of the target user for each point of interest.

[0109] The model parameters are optimized using the stochastic gradient descent algorithm, including the following steps:

[0110] Optimize model parameters using a BPR-based loss function:

[0111]

[0112] Among them, D u represents the check-in set of user u, τ(.) represents the sigmoid function, λ represents the regularization coefficient, and Θ represents the model parameter.

[0113] The present invention also provides a device for recommending points of interest based on user check-in data, comprising: a point of interest graph construction module, used to obtain a user check-in set, and construct a user-point of interest graph and a time-point of interest graph according to the user check-in set; a graph structure feature extraction module, used to extract feature vectors of users and points of interest from the user-point of interest graph, the feature vectors of users and points of interest are used to describe the association between users and points of interest, and to extract feature vectors of time and points of interest from the time-point of interest graph, the feature vectors of time and points of interest are used to describe the association between time and points of interest; a geographical factor evaluation module, used to calculate the geographical factor coefficients of points of interest; a score calculation module, used to calculate the preference score of the target user for each point of interest according to the feature vectors of users and points of interest, the feature vectors of time and points of interest, and the geographical factor coefficients of points of interest; a recommendation module, used to generate a recommendation sequence for the target according to the preference score of the target user for each point of interest.

[0114] The present invention also provides a point of interest recommendation system based on user check-in data, 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 point of interest recommendation method based on user check-in data as described above is implemented.

[0115] The processor and the memory may be connected via a bus or other means.

[0116] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0117] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0118] In addition, an embodiment of the present invention also provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are executed by a processor or controller, for example, by a processor in the above-mentioned terminal embodiment, so that the above-mentioned processor can execute the interest point recommendation method based on user check-in data in the above-mentioned embodiment.

[0119] It will be appreciated by those skilled in the art that all or some of the steps and systems in the disclosed method above may be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that may be used to store desired information and may be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0120] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above-mentioned implementation mode. Technical personnel familiar with the field can also make various equivalent deformations or substitutions without violating the spirit of the present invention. These equivalent deformations or substitutions are all included in the scope defined by the claims of the present invention.

[0121] The specific implementation of the present invention described above does not constitute a limitation on the protection scope of the present invention. Any other corresponding changes and modifications made based on the technical concept of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A method for recommending points of interest based on user check-in data, characterized in that: The following steps are involved: Obtain a user check-in set, and construct a user-interest point graph and a time-interest point graph according to the user check-in set; Extracting feature vectors of users and points of interest from the user-point of interest graph, wherein the feature vectors of users and points of interest are used to describe the association relationship between users and points of interest, and extracting feature vectors of time and points of interest from the time-point of interest graph, wherein the feature vectors of time and points of interest are used to describe the association relationship between time and points of interest; Calculate the geographical factor coefficient of the point of interest; Calculate the target user's preference score for each point of interest based on the feature vectors of the user and the point of interest, the feature vectors of the time and the point of interest, and the geographic factor coefficient of the point of interest; A recommendation sequence for the target is generated according to the target user's preference score for each point of interest.

2. The method for recommending points of interest based on user check-in data according to claim 1, characterized in that: The step of generating a recommendation sequence for the target user according to the target user's preference score for each point of interest comprises the following steps: The interest points are sorted in descending order according to the preference scores, and the first k interest points are taken to form a recommendation sequence for the target user.

3. The method for recommending points of interest based on user check-in data according to claim 1, characterized in that: After the target user's preference score for each point of interest is calculated based on the feature vectors of the user and the point of interest, the feature vectors of the time and the point of interest, and the geographical factor coefficient of the point of interest, the steps include: The model parameters are optimized using the stochastic gradient descent algorithm.

4. The method for recommending points of interest based on user check-in data according to claim 1, characterized in that: The method of calculating the geographical factor coefficient of the point of interest includes the following steps: Get geographic distance function and geographic distance threshold; Calculate a geographic distance coefficient according to the geographic distance function and the geographic distance threshold; The geographic factor coefficient of the point of interest is calculated according to the geographic distance coefficient.

5. The method for recommending points of interest based on user check-in data according to claim 4, characterized in that: The geographic distance coefficient calculated according to the geographic distance function and the geographic distance threshold is expressed as: dis(l,l p ) represents the geographic distance coefficient between the point of interest p and the current location l, geo(l,l p ) represents the geographical distance function between two points calculated by longitude and latitude, d o Indicates the geographic distance threshold, max(.) means taking the maximum value; The geographic factor coefficient of the point of interest calculated according to the geographic distance coefficient is expressed as:

6. The method for recommending points of interest based on user check-in data according to claim 1, characterized in that: The preference score of the target user for each point of interest is calculated based on the feature vectors of the user and the point of interest, the feature vectors of the time and the point of interest, and the geographical factor coefficient of the point of interest, and is expressed as follows: Among them, time is t x , position is l o , user is u i , the point of interest is p j .

7. The method for recommending points of interest based on user check-in data according to claim 3, characterized in that: The model parameters are optimized using the stochastic gradient descent algorithm, including the following steps: Optimize model parameters using a BPR-based loss function: Among them, D u represents the check-in set of user u, τ(.) represents the sigmoid function, λ represents the regularization coefficient, Θ represents the model parameter, The point of interest is p j The preference score of Point of interest p i preference score.

8. A device for recommending points of interest based on user check-in data, characterized in that: include: An interest point graph construction module is used to obtain a user check-in set and construct a user-interest point graph and a time-interest point graph according to the user check-in set; A graph structure feature extraction module, configured to extract feature vectors of users and points of interest from the user-point of interest graph, wherein the feature vectors of users and points of interest are used to describe the association relationship between users and points of interest, and to extract feature vectors of time and points of interest from the time-point of interest graph, wherein the feature vectors of time and points of interest are used to describe the association relationship between time and points of interest; A geographical factor evaluation module is used to calculate the geographical factor coefficient of the point of interest; A score calculation module, used to calculate the target user's preference score for each point of interest based on the feature vectors of the user and the point of interest, the feature vectors of the time and the point of interest, and the geographical factor coefficient of the point of interest; The recommendation module is used to generate a recommendation sequence for the target according to the preference score of the target user for each point of interest.

9. A point of interest recommendation system based on user check-in data, characterized in that: include: 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 method for recommending points of interest based on user check-in data as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the method for recommending points of interest based on user check-in data as described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • Position social network-oriented geographic position regularization-based interest point recommendation method

    CN108460101A

  • Interest point recommendation method and device based on space-time sequence and social embedding ranking

    CN111209475A