Strategy-Aware Tourist Attraction Bundle Recommendation Method

Abstract

This invention provides a strategy-aware method, system, storage medium, and electronic device for recommending bundled tourist attractions, belonging to the field of recommendation system technology. In this invention, firstly, information on the user-attraction set, the user-attraction interaction matrix, and a predefined set of attraction bundles are obtained; secondly, feature representations of users and attractions are obtained; thirdly, an HGT model with a self-attention mechanism is introduced to deeply fuse user preferences and attraction features, obtaining strategy-aware bundle representations and strategy-aware user representations respectively; finally, bundled recommendations are generated based on strategy awareness. This method, when processing tourist attraction recommendation tasks, no longer adheres to a predefined set of attraction bundles, but comprehensively utilizes the multidimensional preference information of the target user and the attribute features of the attractions to generate more personalized and strategic recommendations, ensuring the accuracy of the bundled recommendation results, thereby significantly improving user experience and satisfaction.

Description

Technical Field

[0001] This invention relates to the field of recommendation system technology, and more specifically to a strategy-aware method, system, storage medium, and electronic device for recommending bundled tourist attractions. Background Technology

[0002] With the development of the tourism industry, tourism recommendation systems have gradually become an important tool to help users choose suitable attractions and optimize their travel experience.

[0003] In related technologies, travel recommendation systems mostly employ traditional content-based or collaborative filtering methods. However, these methods still have limitations when faced with complex and ever-changing user needs. Many existing bundled recommendation systems use predefined bundles, which are usually determined based on general trends in big data or operational strategies, failing to meet the highly personalized needs of users. For example, some systems recommend fixed classic travel routes, but these routes may not necessarily match the interests and needs of every user. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] To address the shortcomings of existing technologies, this invention provides a strategy-aware method, system, storage medium, and electronic device for recommending bundled tourist attractions, solving the technical problem that predefined attraction bundles cannot meet users' highly personalized needs.

[0006] (II) Technical Solution

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A strategy-aware method for recommending bundled tourist attractions includes:

[0009] The system acquires a user set and a user information set for each user, a scenic spot set and a scenic spot information set for each scenic spot, a user-scenic spot interaction matrix that records the target user's access history, and a predefined set of scenic spot bundles to construct a bundle-scenic spot interaction graph; wherein any scenic spot bundle is a set consisting of several scenic spots in the scenic spot set.

[0010] Based on the user information set of the target user and the attraction information set of each attraction, the initial feature embedding vectors of the target user and each attraction are obtained respectively.

[0011] Based on the initial feature embedding vectors of the target user and each attraction, and combined with the user-attraction interaction matrix, user token representations and attraction token representations are obtained respectively, and the bundled representation of each attraction is obtained.

[0012] Based on the aforementioned bundled-attraction interaction graph, as well as the user token representation and attraction token representation, an HGT model with a self-attention mechanism is introduced to obtain policy-aware bundle representation and policy-aware user representation, respectively.

[0013] Based on the strategy-aware bundling representation and the strategy-aware user representation, the interaction probability between the target user and each attraction bundling is obtained, so as to obtain the target user's preference value for each attraction in the corresponding attraction bundling.

[0014] The system calculates the target user's preference value for each attraction in all attraction bundles, sorts them, selects the top-ranked attractions, combines them to generate an attraction set, and recommends it to the target user as a bundle.

[0015] Preferably, the process of constructing the bundled-attraction interaction map is as follows:

[0016] The attraction and attraction bundle are considered as two different types of nodes, and the dependency relationship between them is considered as an edge between the attraction node and the bundle node, thus constructing the bundle-attraction interaction graph.

[0017] Preferably, an adaptive graph convolutional network is used to obtain the user token representation and the attraction token representation, respectively, as follows:

[0018]

[0019] in, They represent the target user u respectively m User Token representation, attractions n The attraction token is used for representation; AGCN is an adaptive graph convolutional network; x m ,y n They represent the target user u respectively m and attractions n The initial feature embedding vector;

[0020] Based on the attraction token representations of all attractions in each attraction bundle, obtain the corresponding bundle representation; represented as:

[0021]

[0022] in, This indicates that attractions are bundled together (b). k The binding representation, This indicates that attractions are bundled together (b). k The middle of the o k Each scenic spot Token representation, O k Bundling attractions with b k The number of tourist attractions in China.

[0023] Preferably, the process of acquiring the policy-aware bundled representation includes:

[0024] Introducing an HGT model with L stacked layers, for the l-th layer of the HGT model, based on the bundle-spot interaction graph G, calculate the spot bundle b. k The middle of the o k Each scenic spot Position encoding in the diagram Represented as:

[0025]

[0026] in, This indicates the attraction bundle b obtained from the previous layer of learning. k The latent vector; and These represent attractions on the bundled-attraction interaction map G. Nodes and attractions are bundled together. k The set of neighbors of a node;

[0027] Bundling attractions k Location coding and bundled representation of all attractions Combined, obtain the enhanced binding representation of the l-th layer. Represented as:

[0028]

[0029] in, This represents matrix addition.

[0030] Simplify the self-attention mechanism, retain only the core structure, and compute the attraction bundle b. k Self-attention matrix between in-spots Represented as:

[0031]

[0032] in, and Let be trainable matrices, representing the attraction tokens respectively. The key-value mapping; d represents the embedding dimension; the superscript T represents transpose;

[0033] By combining the self-attention matrix, obtain the attraction bundle b in the l-th layer. k Representation of bundling strategies Represented as:

[0034]

[0035] Where softmax(·) represents the softmax function;

[0036] Perform average pooling on the representation containing the bundling strategy to obtain the bundle b for each attraction. k Strategy-aware binding representation Represented as:

[0037]

[0038] Here, mean_pooling is the average pooling operation.

[0039] Preferably, the process of acquiring the policy-aware user representation includes:

[0040] In layer l, for target user u m The user's historical bundling representation is represented using a representation that includes the bundling strategy, denoted as:

[0041]

[0042] in, Is the target user u m The r-th bundle that has been interacted with, where R is the total number of bundles interacted with by the target user;

[0043] Calculate target user u m Association matrix between its user history and its associated representation Represented as:

[0044]

[0045] in, and Representing user tokens respectively The key-value mapping matrix;

[0046] For the correlation matrix After extracting features from user history and associated representations, and combining them with user token representations... Get target user u m Strategy-aware user representation Represented as:

[0047]

[0048] Where Attention(·) represents the user-bundle association matrix Representation tied to user history Perform interactive modeling; Aggregate(·) represents the feature aggregation function.

[0049] Preferably, based on the policy-aware bundled representation and the policy-aware user representation, the target user u is obtained through an inner product operation. m bundled with attractionsk The interaction probability between them is expressed as:

[0050]

[0051] Where σ(·) is the sigmoid activation function.

[0052] Preferably, the interaction probability is mapped to the target user u using a uniform distribution method. m Bundling attractions with b k Preference values ​​for each attraction Represented as:

[0053]

[0054] A strategy-aware tourist attraction bundling recommendation system includes:

[0055] The information acquisition module is used to acquire a user set and a user information set for each user, a scenic spot set and a scenic spot information set for each scenic spot, a user-scenic spot interaction matrix that records the target user's access history, and a predefined set of scenic spot bundles, in order to construct a bundle-scenic spot interaction graph; wherein, any scenic spot bundle is a set consisting of several scenic spots in the scenic spot set;

[0056] The vector embedding module is used to obtain the initial feature embedding vectors of the target user and each scenic spot based on the user information set of the target user and the scenic spot information set of each scenic spot, respectively.

[0057] The representation acquisition module is used to acquire user token representation and attraction token representation respectively based on the initial feature embedding vector of the target user and each attraction, combined with the user-attraction interaction matrix, and to acquire the bundled representation of each attraction.

[0058] The strategy-aware module is used to introduce an HGT model with a self-attention mechanism based on the bundled-attraction interaction graph, the user token representation, and the attraction token representation, to obtain the strategy-aware bundled representation and the strategy-aware user representation, respectively.

[0059] The preference acquisition module is used to obtain the interaction probability between the target user and each attraction bundle based on the policy-aware bundling representation and the policy-aware user representation, so as to obtain the target user's preference value for each attraction in the corresponding attraction bundle.

[0060] The bundled recommendation module is used to calculate the target user's preference value for each attraction in all attraction bundles, sort them, select the top-ranked attractions, combine them to generate an attraction set as a bundle, and recommend it to the target user.

[0061] A storage medium storing a computer program for policy-aware tourist attraction bundling recommendations, wherein the computer program causes a computer to execute the tourist attraction bundling recommendation method as described above.

[0062] An electronic device, comprising:

[0063] One or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including methods for performing the tourist attraction bundled recommendation method as described above.

[0064] (III) Beneficial Effects

[0065] This invention provides a strategy-aware method, system, storage medium, and electronic device for recommending bundled tourist attractions. Compared with existing technologies, it has the following advantages:

[0066] This invention first acquires user-attraction set information, a user-attraction interaction matrix, and a predefined attraction bundle set; second, it acquires feature representations of users and attractions; third, it introduces an HGT model with a self-attention mechanism to deeply fuse user preferences and attraction features, obtaining strategy-aware bundle representations and strategy-aware user representations respectively; finally, it generates bundled recommendations based on strategy awareness. This method, when handling tourist attraction recommendation tasks, no longer adheres to a predefined attraction bundle set, but comprehensively utilizes the multidimensional preference information of target users and the attribute features of attractions to generate more personalized and strategic recommendations, ensuring the accuracy of bundled recommendation results and thus significantly improving user experience and satisfaction. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 A block diagram illustrating a strategy-aware tourist attraction bundling recommendation method provided in an embodiment of the present invention;

[0069] Figure 2 This is a flowchart of a policy-aware model provided for an embodiment of the present invention. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] This application provides a strategy-aware tourist attraction bundling recommendation method, system, storage medium, and electronic device, which solves the technical problem that predefined attraction bundling combinations cannot meet users' highly personalized needs. It aims to comprehensively consider user preferences, attraction characteristics, and strategy-aware factors when processing attraction bundling recommendation tasks in tourism scenarios, thereby generating more targeted and highly satisfactory tourism recommendation solutions, improving the accuracy of recommendation results and user experience.

[0072] The technical solution in this application is to solve the above-mentioned technical problems, and the general idea is as follows:

[0073] In this embodiment of the invention, firstly, the set information of users and attractions, the user-attraction interaction matrix, and a predefined set of attraction bundles are obtained; secondly, the feature representations of users and attractions are obtained; thirdly, an HGT model with a self-attention mechanism is introduced to deeply fuse user preferences and attraction features, obtaining strategy-aware bundle representations and strategy-aware user representations respectively; finally, bundled recommendations are generated based on strategy awareness. This method, when processing tourist attraction recommendation tasks, no longer adheres to a predefined set of attraction bundles, but comprehensively utilizes the multidimensional preference information of target users and the attribute features of attractions to generate more personalized and strategic recommendations, ensuring the accuracy of bundled recommendation results, thereby significantly improving user experience and satisfaction.

[0074] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0075] Example 1:

[0076] like Figure 1 As shown, this embodiment of the invention provides a strategy-aware method for bundling and recommending tourist attractions, including:

[0077] S1. Obtain a user set and a user information set for each user, a scenic spot set and a scenic spot information set for each scenic spot, a user-scenic spot interaction matrix that records the target user's access history, and a predefined scenic spot bundle set to construct a bundle-scenic spot interaction graph; wherein, any scenic spot bundle is a set consisting of several scenic spots in the scenic spot set.

[0078] S2. Based on the user information set of the target user and the attraction information set of each attraction, obtain the initial feature embedding vectors of the target user and each attraction respectively.

[0079] S3. Based on the initial feature embedding vectors of the target user and each attraction, and combined with the user-attraction interaction matrix, obtain the user token representation and attraction token representation respectively, and obtain the binding representation of each attraction.

[0080] S4. Based on the bundled-attraction interaction graph, and the user token representation and attraction token representation, introduce an HGT model with a self-attention mechanism to obtain the strategy-aware bundled representation and the strategy-aware user representation respectively.

[0081] S5. Based on the strategy-aware bundling representation and the strategy-aware user representation, obtain the interaction probability between the target user and each attraction bundling, so as to obtain the target user's preference value for each attraction in the corresponding attraction bundling.

[0082] S6. Calculate the target user's preference value for each attraction in all attraction bundles, sort them, select the top-ranked attractions, combine them to generate an attraction set as a bundle, and recommend it to the target user.

[0083] In processing tourist attraction recommendation tasks, this invention no longer adheres to a predefined set of attraction bundles. Instead, it comprehensively utilizes the multidimensional preference information of the target user and the attribute features of the attractions to generate more personalized and strategic recommendations, ensuring the accuracy of the bundled recommendation results and thus significantly improving user experience and satisfaction.

[0084] The following will detail each step of the above solution:

[0085] In step S1, a user set and a user information set for each user, a scenic spot set and a scenic spot information set for each scenic spot, a user-scenic spot interaction matrix that records the target user's access history, and a predefined scenic spot bundle set are obtained to construct a bundle-scenic spot interaction graph; wherein, any scenic spot bundle is a set consisting of several scenic spots in the scenic spot set.

[0086] Specifically, this step obtains the following information:

[0087] (1) The user set, denoted as U = {u1, u2, ..., u...} m ,…,u M}, where u m Let m represent the m-th user, where M is the total number of users.

[0088] In addition, for target user u m(Assuming the m-th user is the target user), collect their user ID and preferences such as travel style, budget, and attraction reviews to construct a user information set. in Indicates user u m The value of the i-th piece of information, where I is the total number of user information.

[0089] (2) The set of attractions is denoted as S = {s1, s2, ..., s}. n ,…,s N}, where s n Let N represent the nth attraction, where N is the total number of attractions.

[0090] In addition, for attractions s n Collect attraction IDs and attribute information such as location, type, and rating to construct an attraction information set. in Indicates the attraction s n The value of the j-th piece of information, where J is the total number of attraction information.

[0091] (3) Record the user-attraction interaction matrix of the target user's access history, denoted as Where X mn =1 indicates the target user u m Visited attractions n Otherwise X mn =0.

[0092] (4) A predefined set of attraction bundles, denoted as B = {b1, b2, ..., b} k ,…,b K}, where b k This represents the k-th bundle, where K is the total number of bundles. (Scenic spot bundle b) k It is a set consisting of several attractions from the attraction set S, denoted as . in This indicates that attractions are bundled together (b). k The oth k One attraction, O k Bundling attractions with b k The number of tourist attractions in China.

[0093] Based on this, this step constructs a bundled attraction interaction diagram:

[0094] Treat attractions and attraction bundles as two different types of nodes, and the dependency relationship between them as an edge between the attraction node and the bundle node, constructing a bundle-attraction interaction graph G.

[0095] In step S2, based on the user information set of the target user and the attraction information set of each attraction, the initial feature embedding vectors of the target user and each attraction are obtained respectively.

[0096] This step transforms user and attraction information into trainable vectors, and uses equations (1) and (2) to obtain the target user u. m and attractions n initial feature embedding vector and

[0097] x m =f embed (P m (1)

[0098] y n =f embed (Q n (2)

[0099] Among them, f embed (·) represents the embedding mapping function, and d represents the embedding dimension.

[0100] In step S3, based on the target user and the initial feature embedding vector of each attraction, and combined with the user-attraction interaction matrix, user token representation and attraction token representation are obtained respectively, and the bundled representation of each attraction is obtained.

[0101] This step treats the attractions within the bundle as tokens and embeds them into the model to obtain a representation of the attraction bundles for subsequent policy-aware processing. Specifically, this includes:

[0102] S31. Combining the user-attraction interaction matrix, an adaptive graph convolutional network is used to obtain the optimized user u. m and attractions n Token representation, expressed as:

[0103]

[0104] in, They represent the target user u respectively m User Token representation, attractions n The attraction token is represented by the AGCN, which is an adaptive graph convolutional network.

[0105] S32. Based on the attraction token representations of all attractions in each attraction bundle, obtain the corresponding bundle representation; represented as:

[0106]

[0107] in, This indicates that attractions are bundled together (b). k The binding representation, This indicates that attractions are bundled together (b). kThe middle of the o k Each scenic spot Token representation.

[0108] In step S4, based on the bundled-attraction interaction graph, the user token representation, and the attraction token representation, an HGT model with a self-attention mechanism is introduced to obtain the policy-aware bundled representation and the policy-aware user representation, respectively.

[0109] This invention constructs an L-layer HGT model to model bundles and users within a graph structure, capturing the complex relationships between them and obtaining policy-aware representations of bundles and users.

[0110] like Figure 2 As shown, Figure 2 A flowchart based on the policy-aware model provided in an embodiment of the present invention is given.

[0111] Reference Figure 2 The process of obtaining the strategy-aware binding representation in this step specifically includes:

[0112] S41. Each attraction bundle implicitly contains strategy information for selecting a particular attraction. By introducing statistical relationships between attractions in the bundle, a graph-based location encoder is used to reflect the prior knowledge of the predefined bundling strategy. Specifically, for the l-th (1≤l≤L) layer of the HGT model, based on the bundling-attraction interaction graph G, the attraction bundle b is calculated. k The middle of the o k Each scenic spot Position encoding in the diagram Represented as:

[0113]

[0114] in, This indicates the attraction bundle b obtained from the previous layer of learning. k The potential vector, when l=1, is obtained by random initialization; and These represent attractions on the bundled-attraction interaction map G. Nodes and attractions are bundled together. k The set of neighbors of a node.

[0115] S42, Bundling attractions together k Location coding and bundled representation of all attractions Combined, obtain the enhanced binding representation of the l-th layer. Represented as:

[0116]

[0117] in, This represents matrix addition.

[0118] S43. After obtaining the bundle representation, a self-attention mechanism is used to learn the associations between points within the bundle for bundling strategy modeling. Considering issues such as sparsity, the self-attention mechanism is simplified by removing the feedforward network (FFN), multi-head attention scoring, and hierarchical normalization, retaining only the core structure, and calculating the point bundle b. k Self-attention matrix between in-spots Represented as:

[0119]

[0120] in, and Let be trainable matrices, representing the attraction tokens respectively. The key-value mapping; the superscript T indicates transpose.

[0121] S44. Combine the self-attention matrix to obtain the attraction bundle b in the l-th layer. k Representation of bundling strategies Represented as:

[0122]

[0123] Here, softmax(·) represents the softmax function.

[0124] S45. Perform average pooling on the representation containing the bundling strategy to obtain the bundling b for each attraction. k Strategy-aware binding representation Represented as:

[0125]

[0126] Here, mean_pooling is the average pooling operation.

[0127] Similarly, refer to Figure 2 The specific steps of obtaining user representations for policy awareness in this process include:

[0128] S46. In the l-th layer, for the target user u m The user's historical bundling representation is represented using a representation that includes the bundling strategy, denoted as:

[0129]

[0130] in, Is the target user u m The r-th bundle that has been interacted with, where R is the total number of bundles interacted with by the target user.

[0131] S47. Calculate the target user um Association matrix between its user history and its associated representation Represented as:

[0132]

[0133] in, and Representing user tokens respectively The key-value mapping matrix;

[0134] S48. On the correlation matrix After extracting features from user history and associated representations, and combining them with user token representations... Get target user u m Strategy-aware user representation Represented as:

[0135]

[0136] Where Attention(·) represents the user-bundle association matrix Representation tied to user history Perform interactive modeling; Aggregate(·) represents the feature aggregation function.

[0137] In step S5, based on the policy-aware bundling representation and the policy-aware user representation, the interaction probability between the target user and each attraction bundling is obtained, so as to obtain the target user's preference value for each attraction in the corresponding attraction bundling.

[0138] In this step, such as Figure 2 As shown, based on the policy-aware bundled representation and the policy-aware user representation, the target user u is obtained through the inner product operation. m bundled with attractions k The interaction probability between them is expressed as:

[0139]

[0140] Where σ(·) is the sigmoid activation function.

[0141] Building on this, this step then uses a uniform distribution method to map the interaction probability to the target user um's preference value for each attraction in the attraction bundle bk. Represented as:

[0142]

[0143] In step S6, the target user's preference value for each attraction in all attraction bundles is calculated, and after sorting, several top-ranked attractions are selected and combined to generate an attraction set as a bundle for recommendation to the target user.

[0144] In this step, for the target user u m The preference values ​​of each attraction in all attraction bundles are sorted in non-ascending order, and the attractions corresponding to the top 10 preference values ​​are selected to generate a set of attractions as a bundle for the target user u. m Recommendations will be made.

[0145] Thus, this embodiment of the invention completes the entire process of the strategy-aware tourist attraction bundling recommendation method.

[0146] Example 2:

[0147] This invention provides a strategy-aware tourist attraction bundling recommendation system, comprising:

[0148] The information acquisition module is used to acquire a user set and a user information set for each user, a scenic spot set and a scenic spot information set for each scenic spot, a user-scenic spot interaction matrix that records the target user's access history, and a predefined set of scenic spot bundles, in order to construct a bundle-scenic spot interaction graph; wherein, any scenic spot bundle is a set consisting of several scenic spots in the scenic spot set;

[0149] The vector embedding module is used to obtain the initial feature embedding vectors of the target user and each scenic spot based on the user information set of the target user and the scenic spot information set of each scenic spot, respectively.

[0150] The representation acquisition module is used to acquire user token representation and attraction token representation respectively based on the initial feature embedding vector of the target user and each attraction, combined with the user-attraction interaction matrix, and to acquire the bundled representation of each attraction.

[0151] The strategy-aware module is used to introduce an HGT model with a self-attention mechanism based on the bundled-attraction interaction graph, the user token representation, and the attraction token representation, to obtain the strategy-aware bundled representation and the strategy-aware user representation, respectively.

[0152] The preference acquisition module is used to obtain the interaction probability between the target user and each attraction bundle based on the policy-aware bundling representation and the policy-aware user representation, so as to obtain the target user's preference value for each attraction in the corresponding attraction bundle.

[0153] The bundled recommendation module is used to calculate the target user's preference value for each attraction in all attraction bundles, sort them, select the top-ranked attractions, combine them to generate an attraction set as a bundle, and recommend it to the target user.

[0154] Example 3:

[0155] This invention provides a storage medium storing a computer program for policy-aware tourist attraction bundling recommendations, wherein the computer program causes a computer to execute the tourist attraction bundling recommendation method as described in Embodiment 1.

[0156] Example 4:

[0157] This invention provides an electronic device, comprising:

[0158] One or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including methods for performing tourist attraction bundled recommendation methods as described in Example 1.

[0159] It is understood that the strategy-aware tourist attraction bundling recommendation system, storage medium, and electronic device provided in the embodiments of the present invention correspond to the strategy-aware tourist attraction bundling recommendation method provided in the embodiments of the present invention. The explanations, examples, and beneficial effects of the relevant content can be referred to the corresponding parts of the tourist attraction bundling recommendation method, and will not be repeated here.

[0160] In summary, compared with existing technologies, it has the following beneficial effects:

[0161] 1. Strategy Awareness: By introducing adaptive graph convolutional networks and self-attention mechanisms, this invention can dynamically adjust the bundling strategy to better capture the complex relationships between attraction bundles, achieving more accurate recommendation combinations. This strategy awareness helps the recommendation system flexibly adapt to and optimize recommendation performance in complex tourism scenarios.

[0162] 2. Efficient computational model: By combining a hierarchical graph transformer (HGT) and a self-attention mechanism, this invention can efficiently process large-scale tourist attraction and user data, ensuring the real-time performance and response speed of the system while generating recommendations.

[0163] 3. Highly Personalized Recommendations: The system accurately models users' bundled preferences, generating bundled recommendations of tourist attractions that match users' interests and preferences, thereby significantly improving user experience and satisfaction.

[0164] 4. Flexible scalability: It is not only applicable to the recommendation task of tourist attractions, but can also be flexibly extended to other types of bundled recommendation tasks, such as product recommendation, video recommendation, etc., thus having broad application potential and market adaptability.

[0165] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0166] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A strategy-aware method for recommending bundled tourist attractions, characterized in that, include: The system acquires a user set and a user information set for each user, a scenic spot set and a scenic spot information set for each scenic spot, a user-scenic spot interaction matrix that records the target user's access history, and a predefined set of scenic spot bundles to construct a bundle-scenic spot interaction graph; wherein any scenic spot bundle is a set consisting of several scenic spots in the scenic spot set. Based on the user information set of the target user and the attraction information set of each attraction, the initial feature embedding vectors of the target user and each attraction are obtained respectively. Based on the initial feature embedding vectors of the target user and each attraction, and combined with the user-attraction interaction matrix, user token representations and attraction token representations are obtained respectively, and the bundled representation of each attraction is obtained. Based on the aforementioned bundled-attraction interaction graph, as well as the user token representation and attraction token representation, an HGT model with a self-attention mechanism is introduced to obtain policy-aware bundle representation and policy-aware user representation, respectively. Based on the strategy-aware bundling representation and the strategy-aware user representation, the interaction probability between the target user and each attraction bundling is obtained, so as to obtain the target user's preference value for each attraction in the corresponding attraction bundling. The system calculates the target user's preference value for each attraction in all attraction bundles, sorts them, selects the top-ranked attractions, combines them to generate an attraction set, and recommends it to the target user as a bundle.

2. The method for bundling and recommending tourist attractions as described in claim 1, characterized in that, The process of constructing the bundled-attraction interaction map is as follows: The attraction and attraction bundle are considered as two different types of nodes, and the dependency relationship between them is considered as an edge between the attraction node and the bundle node, thus constructing the bundle-attraction interaction graph.

3. The method for bundling and recommending tourist attractions as described in claim 1, characterized in that, An adaptive graph convolutional network is used to obtain user token representations and attraction token representations, respectively, as follows: in, They represent the target user u respectively m User Token Representation, Attractions n The attraction token is used for representation; AGCN is an adaptive graph convolutional network; x m y n They represent the target user u respectively m and attractions n The initial feature embedding vector; X is the user-attraction interaction matrix; Based on the attraction token representations of all attractions in each attraction bundle, obtain the corresponding bundle representation; represented as: in, This indicates that attractions are bundled together (b). k The binding representation, This indicates that attractions are bundled together (b). k The middle of the o k Each scenic spot Token representation, O k Bundling attractions with b k The number of tourist attractions in China.

4. The method for bundling and recommending tourist attractions as described in claim 3, characterized in that, The process of acquiring the policy-aware bundled representation includes: Introducing an HGT model with L stacked layers, for the l-th layer of the HGT model, based on the bundle-spot interaction graph G, calculate the spot bundle b. k The middle of the o k Each scenic spot Position encoding in the diagram Represented as: in, This indicates the attraction bundle b obtained from the previous layer of learning. k The latent vector; and These represent attractions on the bundled-attraction interaction map G. Nodes and attractions are bundled together. k The set of neighbors of a node; Bundling attractions k Location coding and bundled representation of all attractions Combined, obtain the enhanced binding representation of the l-th layer. Represented as: in, This represents matrix addition. Simplify the self-attention mechanism, retain only the core structure, and compute the attraction bundle b. k Self-attention matrix between in-spots Represented as: in, and Let be trainable matrices, representing the attraction tokens respectively. The key-value mapping; d represents the embedding dimension; the superscript T represents transpose; By combining the self-attention matrix, obtain the attraction bundle b in the l-th layer. k Representation of bundling strategies Represented as: Where softmax(·) represents the softmax function; Perform average pooling on the representation containing the bundling strategy to obtain the bundle b for each attraction. k Strategy-aware binding representation Represented as: Here, mean_pooling is the average pooling operation.

5. The method for bundling and recommending tourist attractions as described in claim 4, characterized in that, The process of acquiring the strategy-aware user representation includes: In the l-th layer, for the target user u m The user's historical bundling representation is represented using a representation that includes the bundling strategy, denoted as: in, Is the target user u m The r-th bundle that has been interacted with, where R is the total number of bundles interacted with by the target user; Calculate target user u m Association matrix between its user history and its associated representation Represented as: in, and Representing user tokens respectively The key-value mapping matrix; the correlation matrix After extracting features from user history and associated representations, and combining them with user token representations... Get target user u m Strategy-aware user representation Represented as: Where Attention(·) represents the user-bundling association matrix Representation tied to user history Perform interactive modeling; Aggregate(·) represents the feature aggregation function.

6. The method for bundling and recommending tourist attractions as described in claim 5, characterized in that, Based on the aforementioned policy-aware bundle representation and policy-aware user representation, the target user u is obtained through an inner product operation. m bundled with attractions k The interaction probability between them is expressed as: Where σ(·) is the sigmoid activation function.

7. The method for bundling and recommending tourist attractions as described in claim 6, characterized in that, The interaction probability is mapped to the target user u using a uniform distribution method. m Bundling attractions with b k Preference values ​​for each attraction Represented as:

8. A strategy-aware tourist attraction bundling recommendation system, characterized in that, include: The information acquisition module is used to acquire a user set and a user information set for each user, a scenic spot set and a scenic spot information set for each scenic spot, a user-scenic spot interaction matrix that records the target user's access history, and a predefined set of scenic spot bundles, in order to construct a bundle-scenic spot interaction graph; wherein, any scenic spot bundle is a set consisting of several scenic spots in the scenic spot set; The vector embedding module is used to obtain the initial feature embedding vectors of the target user and each scenic spot based on the user information set of the target user and the scenic spot information set of each scenic spot, respectively. The representation acquisition module is used to acquire user token representation and attraction token representation respectively based on the initial feature embedding vector of the target user and each attraction, combined with the user-attraction interaction matrix, and to acquire the bundled representation of each attraction. The strategy-aware module is used to introduce an HGT model with a self-attention mechanism based on the bundled-attraction interaction graph, the user token representation, and the attraction token representation, to obtain the strategy-aware bundled representation and the strategy-aware user representation, respectively. The preference acquisition module is used to obtain the interaction probability between the target user and each attraction bundle based on the policy-aware bundling representation and the policy-aware user representation, so as to obtain the target user's preference value for each attraction in the corresponding attraction bundle. The bundled recommendation module is used to calculate the target user's preference value for each attraction in all attraction bundles, sort them, select the top-ranked attractions, combine them to generate an attraction set as a bundle, and recommend it to the target user.

9. A storage medium, characterized in that, It stores a computer program for strategy-aware tourist attraction bundling recommendations, wherein the computer program causes a computer to perform the tourist attraction bundling recommendation method as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, include: One or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including methods for performing tourist attraction bundled recommendation as described in any one of claims 1 to 7.

Images