Shopping center shop recommendation method and device, equipment and medium

Abstract

The invention relates to a shopping center shop recommendation method and device, equipment and a medium, and the method comprises the steps: predicting interested shops of a target old user in a current time period through employing a long and short-term memory network model trained by user behavior time sequence data, adopting a preset collaborative filtering algorithm to calculate and determine a second similarity between the target old user and other users according to the interested shops of the target old user in the current time period, so as to determine neighbor users of the target old user; predicting an estimated score value of a target old user for an unscored shop according to the score value of the neighbor user for each shop, and calculating and determining a second recommendation degree corresponding to each shop according to the estimated score value, the actual distance between the target old user and the shop and the comprehensive recommendation value, and displaying each shop to a user interface of the target old user according to the second recommendation degree from high to low so as to complete shop recommendation of the shopping center. According to the invention, the traffic and user stickiness of a shopping center can be improved.

Description

Technical Field

[0001] The present application relates to the field of commodity recommendation, and in particular to a method for recommending stores in a shopping mall, a corresponding device, an electronic device, and a computer-readable storage medium. Background Art

[0002] Under the background of rapid development of information technology, recommendation systems and path planning algorithms play a vital role in helping users make decisions and process large amounts of information. Nevertheless, these technologies still face some challenges in practical applications.

[0003] The cold start problem is particularly prominent for new users because they lack historical interaction data, which makes it difficult for the system to build an accurate user profile and thus provide personalized recommendations. In addition, the interests of new users may change rapidly in a short period of time, which requires the recommendation system to quickly adapt to these changes to provide timely recommendations. At the same time, due to the lack of sufficient interaction data, the interaction between new users and items is very sparse, which makes it more difficult to capture the user's true preferences.

[0004] For old users, as the number of users and the types of items increase, the interactions between users and items become increasingly sparse, resulting in incomplete data. This sparsity makes it difficult for the system to accurately capture the true preferences of users. In addition, the interests of old users may change over time, environment, and personal experience, and the recommendation system needs to be able to update in real time to respond to these changes. As the amount of data increases, the computational complexity of the algorithm also increases, which not only affects the response speed of the system, but also limits its scalability.

[0005] In terms of path planning, the system needs to balance multiple objectives such as shortest time, lowest cost and highest comfort, which increases the complexity of the optimization process. Obstacles and path changes in the actual environment require the path planning algorithm to be updated in real time to adapt to these changes. In addition, traditional path planning algorithms may not accurately reflect the path and distance actually walked by the user, and more accurate algorithms are needed to provide optimized paths.

[0006] To sum up, in the existing technology, the interaction between new users and items is very sparse, which makes it more difficult to capture the user's true preferences. For old users, as the number of users and the types of items increase, the interaction between users and items becomes increasingly sparse, making it difficult for the system to accurately capture the user's true preferences. The applicant has made corresponding explorations to solve this problem. Summary of the invention

[0007] The purpose of the present application is to solve the above-mentioned problems and to provide a shopping mall store recommendation method, corresponding device, electronic device and computer-readable storage medium.

[0008] In order to meet the various objectives of this application, this application adopts the following technical solutions:

[0009] A method for recommending stores in a shopping mall proposed to meet one of the purposes of this application includes:

[0010] Obtain the number of store reviews and the store comprehensive score corresponding to each store in the shopping mall, determine the store popularity value corresponding to the store according to the number of store reviews and the store comprehensive score, determine the store popularity corresponding to the store according to the store popularity value, the store comprehensive score and the number of new reviews of the store within a preset time period, and determine the comprehensive recommendation value corresponding to each store according to the store popularity value and the store popularity;

[0011] In response to an instruction to recommend a store to a target new user, obtain a user interest category tag of the target new user and a store category tag corresponding to each store, calculate and determine a first similarity between the user interest category tag and the store category tag, determine a first recommendation degree corresponding to each store according to the comprehensive recommendation value, the first similarity, and an actual distance between the target new user and the store, and display each store in a descending order according to the first recommendation degree to a user interface of the target new user;

[0012] In response to an instruction to recommend a store to an old user, the user behavior time series data of the target old user is obtained, wherein the user behavior time series data includes historical consumption records and user consumption levels, the historical consumption records represent information about stores visited and consumed by the user in different time periods, and the user consumption level is obtained by the user's consumption frequency in different stores and the price range of the stores;

[0013] A long short-term memory network model trained with the user behavior time series data is used to predict the target old user's stores of interest in the current period, and a preset collaborative filtering algorithm is used to calculate and determine the second similarity between the target old user and other users based on the stores of interest to the target old user in the current period, so as to determine the neighbor users of the target old user;

[0014] The estimated ratings of the target old user for the unrated stores are predicted based on the ratings of the neighbor users for each store. The second recommendation degree corresponding to each store is calculated and determined based on the estimated ratings, the actual distance between the target old user and the store, and the comprehensive recommendation value. Each store is displayed to the user interface of the target old user from high to low according to the second recommendation degree, so as to complete the store recommendation of the shopping mall.

[0015] Optionally, the store popularity value corresponding to the store is determined according to the number of store reviews and the store comprehensive score. The calculation formula of the store popularity value is:

[0016]

[0017] Among them, Hotness represents the store popularity value corresponding to each store, R1 represents the number of store reviews received by the store, and r represents the comprehensive score of the store;

[0018] The store popularity corresponding to the store is determined according to the store popularity value, the store comprehensive score and the number of new reviews of the store within a preset time range. The calculation formula of the store popularity is:

[0019] P=α·Hotness+β·S+γ·ΔN,

[0020] Among them, P represents the store popularity corresponding to each store, S represents the comprehensive score of the store, ΔN represents the number of new reviews of the store within a preset time range, and the preset time range is within the past month, and α, β and γ represent weight coefficients;

[0021] The comprehensive recommendation value corresponding to each store is determined according to the store heat value and the store popularity. The calculation formula of the comprehensive recommendation value is:

[0022] R = λ·P+(1-λ)·Hotness,

[0023] Among them, R is the comprehensive recommendation value corresponding to each store, and λ is the weight coefficient, which is used to balance the importance of store popularity and store heat value.

[0024] Optionally, a first similarity between the user's interested category label and the store category label is calculated and determined, and the calculation formula of the first similarity is:

[0025]

[0026] Among them, Jaccard similarity represents the first similarity, A is the user's interested category label set, B is the store category label set, and the higher the Jaccard similarity, the greater the user's interest in the store;

[0027] The first recommendation degree corresponding to each store is determined according to the comprehensive recommendation value, the first similarity, and the actual distance between the target new user and the store. The calculation formula of the first recommendation degree is:

[0028]

[0029] Among them, ω1 represents the weight of the comprehensive recommendation value, ω2 represents the weight of the Jaccard similarity, ω3 represents the weight of the actual distance between the target new user and the store, and R represents the comprehensive recommendation value corresponding to each store.

[0030] Optionally, the calculation formula for the user consumption level is:

[0031]

[0032] Among them, T i represents the number of times or frequency of consumption by the user in the i-th store, P i represents the price range of the store, and X represents the user's consumption level.

[0033] Optionally, a preset collaborative filtering algorithm is used to calculate and determine the second similarity between the target old user and other users according to the stores that the target old user is interested in during the current period. The calculation formula of the second similarity is expressed as:

[0034]

[0035] Among them, sim(u,v) represents the second similarity, u represents the vector of the target old user, v represents the vector of other users, u and v contain information about the user's interest in the store and consumption level, θ represents the angle between user vectors, and the value range of cosine similarity cos(θ) is [0,1]. The larger the value, the more similar other users are to the target old user.

[0036] Optionally, the estimated ratings of the target old users for the unrated stores are predicted based on the ratings of the neighbor users for each store, and the calculation formula for the estimated ratings is:

[0037]

[0038] in, is the estimated rating value, which represents the predicted rating of the target old user u for the store i, N represents the set of neighbor users with the highest similarity to the target old user u, ω uv represents the similarity between the target old user u and the neighbor user v, r vi represents the rating of store i by neighbor user v;

[0039] According to the estimated rating value, the actual distance between the target old user and the store, and the comprehensive recommendation value, the second recommendation corresponding to each store is calculated and determined, and the calculation formula of the second recommendation is:

[0040]

[0041] Among them, α, β and γ represent weight coefficients, and R represents the comprehensive recommendation value corresponding to each store.

[0042] Optionally, the step of determining the actual distance between the user and the store includes:

[0043] Obtain a shopping mall map to determine the starting location of a target new user or target old user and multiple target store locations;

[0044] An improved Dijkstra algorithm is used to perform path planning based on the starting point location and multiple target store locations to determine the actual distance from the current location of the target new user or the target old user to each target store.

[0045] A shopping mall store recommendation device provided for another purpose of the present application includes:

[0046] A comprehensive recommendation value determination module is configured to obtain the number of store reviews and the store comprehensive score corresponding to each store in the shopping mall, determine the store popularity value corresponding to the store according to the number of store reviews and the store comprehensive score, determine the store popularity corresponding to the store according to the store popularity value, the store comprehensive score and the number of new reviews of the store within a preset time period, and determine the comprehensive recommendation value corresponding to each store according to the store popularity value and the store popularity;

[0047] A new user store recommendation module is configured to respond to an instruction to recommend a store to a target new user, obtain a user interest category tag of the target new user and a store category tag corresponding to each store, calculate and determine a first similarity between the user interest category tag and the store category tag, determine a first recommendation degree corresponding to each store according to the comprehensive recommendation value, the first similarity, and an actual distance between the target new user and the store, and display each store in a descending order according to the first recommendation degree to a user interface of the target new user;

[0048] A user behavior data acquisition module, configured to respond to an instruction to recommend a store to an old user, and acquire user behavior time series data of a target old user, wherein the user behavior time series data includes historical consumption records and user consumption levels, the historical consumption records represent information about stores visited and consumed by users in different time periods, and the user consumption level is obtained by the consumption frequency of users in different stores and the price range of the stores;

[0049] A neighbor user determination module is configured to use a long short-term memory network model trained with the user behavior time series data to predict the target old user's stores of interest in the current period, and use a preset collaborative filtering algorithm to calculate and determine the second similarity between the target old user and other users based on the stores of interest to the target old user in the current period, so as to determine the neighbor users of the target old user;

[0050] The old user store recommendation module is configured to predict the estimated rating value of the target old user for the unrated stores based on the rating values ​​of the neighbor users for each store, calculate and determine the second recommendation degree corresponding to each store based on the estimated rating value, the actual distance between the target old user and the store, and the comprehensive recommendation value, and display each store to the user interface of the target old user from high to low according to the second recommendation degree, so as to complete the store recommendation of the shopping mall.

[0051] An electronic device provided to meet another purpose of the present application includes a central processing unit and a memory, wherein the central processing unit is used to call and run a computer program stored in the memory to execute the steps of the shopping mall store recommendation method described in the present application.

[0052] A computer-readable storage medium is provided to meet another purpose of the present application, which stores a computer program implemented according to the shopping mall store recommendation method in the form of computer-readable instructions. When the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

[0053] Compared with the prior art, the present application aims at the following problems: the interaction between new users and items is very sparse in the prior art, which makes it more difficult to capture the real preferences of users. For old users, as the number of users and the types of items increase, the interaction between users and items becomes increasingly sparse, making it difficult for the system to accurately capture the real preferences of users. The present application includes but is not limited to the following beneficial effects:

[0054] First, the system can accurately calculate the actual walking distance of users to each store through the improved Dijkstra algorithm, taking into account the actual walking path of users, rather than just the traditional distance measurement (such as Manhattan distance or Euclidean distance). This precise path planning greatly improves the navigation experience of users in shopping malls, provides users with the best walking route, and further optimizes the shopping process.

[0055] Secondly, this application solves the cold start problem for new users. In the new user recommendation strategy, by combining the user-selected tags with the characteristics of popular stores in the current shopping mall, the system can effectively alleviate the cold start problem. The Jaccard similarity coefficient is used to analyze the relationship between the category tags selected by the user and the store category tags, so that new users can quickly find stores that meet their potential interests.

[0056] Third, for old users, the system uses an LSTM neural network model to analyze the user's historical behavior data and accurately capture the user's interest characteristics. Combined with the clustering method of subdivided categories, the sparsity problem is reduced, and the accuracy of recommendations and the processability of data are improved. By weighted calculation of the weight factors in the user-item matrix, the system can more realistically reflect the user's preferences for each store and improve the personalization of recommendations.

[0057] Fourthly, this application uses the user similarity score prediction method to enable the system to more comprehensively predict the user's interest in a store. When a user lacks direct interaction records with certain stores, the system can predict the target user's potential preferences through the scores and feedback of neighboring users. This weighted prediction method based on neighbor users effectively fills the gaps in the user-item matrix and further improves the accuracy and diversity of recommendations.

[0058] Fifth, at the core of the recommendation algorithm, the system comprehensively considers data from multiple dimensions (such as store ratings, number of reviews, new reviews, etc.), and calculates the popularity and heat value of the store through a weighted average model. This multi-dimensional comprehensive calculation ensures the accuracy and timeliness of the recommendation results, making the recommended content more in line with the interests and needs of users.

[0059] Sixth, by collecting statistics on the price range and visit frequency of users entering the store, the system can more accurately build user portraits, analyze user consumption levels, and provide a more reliable basis for personalized recommendations. The similarity calculation between users further enhances the system's recommendation capabilities, thereby ensuring that each user can receive the recommendations that best meet their needs.

[0060] Seventh, this application improves user satisfaction and shopping experience. Combined with accurate route planning, real-time updates and personalized recommendations, the system can greatly improve the user's shopping experience, reduce unnecessary time waste, and help users quickly find stores that match their interests. Overall, the recommendation system can significantly improve user satisfaction and increase the flow and user stickiness of shopping malls.

[0061] In summary, this application implements a recommendation system that comprehensively utilizes user behavior data, store characteristics, and actual distance. Through the combination of deep learning models and multiple algorithms, it provides users with personalized store recommendations, significantly improving user experience and customer satisfaction. This technical solution not only has good practicality, but also contributes to the construction of future smart shopping navigation systems, adding more convenience and fun to users' shopping experience. Combining positioning technology, real-time data analysis, and personalized recommendation algorithms, this application shows broad application prospects and can provide strong support for future smart retail and shopping center management. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0063] Figure 1 A schematic diagram of the flow of a shopping mall store recommendation method in an embodiment of the present application;

[0064] Figure 2 This is a schematic diagram of the process of the broken line turning shopping mall route planning diagram in the embodiment of the present application;

[0065] Figure 3 This is a schematic diagram of the process of the circular shopping mall route planning diagram in the embodiment of the present application;

[0066] Figure 4 This is a schematic diagram of the process of the polygonal shopping mall path planning diagram in the embodiment of the present application;

[0067] Figure 5 This is a schematic diagram of the comparison of the training time of the RNN model, LSTM model and GRU model in the embodiment of the present application;

[0068] Figure 6 A schematic diagram showing the comparison of the inference time among the RNN model, the LSTM model and the GRU model in the embodiment of the present application;

[0069] Figure 7 A schematic diagram showing the performance comparison of the RNN model, LSTM model and GRU model in terms of training loss in an embodiment of the present application;

[0070] Figure 8 This is a schematic diagram of the performance comparison of the RNN model, LSTM model and GRU model in terms of verification loss in the embodiment of the present application;

[0071] Fig. 9 A schematic diagram showing the performance comparison of the RNN model, LSTM model and GRU model in terms of accuracy in the embodiment of the present application;

[0072] Fig.10This is a functional block diagram of a device for recommending stores in a shopping mall in an embodiment of the present application;

[0073] Fig.11 It is a schematic diagram of the structure of the computer device in the embodiment of the present application. DETAILED DESCRIPTION

[0074] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be interpreted as limiting the present application.

[0075] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0076] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with those in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as here.

[0077] It will be understood by those skilled in the art that the "client", "terminal" and "terminal device" used herein include both devices with wireless signal receivers, which are devices with only wireless signal receivers without transmission capabilities, and devices with receiving and transmitting hardware, which are devices with receiving and transmitting hardware capable of two-way communication over a two-way communication link. Such devices may include: cellular or other communication devices such as personal computers, tablet computers, which have single-line displays or multi-line displays or cellular or other communication devices without multi-line displays; PCS (Personal Communications Service, personal communication system), which can combine voice, data processing, fax and / or data communication capabilities; PDA (Personal Digital Assistant, personal digital assistant), which may include a radio frequency receiver, pager, Internet / intranet access, web browser, notepad, calendar and / or GPS (Global Positioning System, global positioning system) receiver; conventional laptop and / or palmtop computers or other devices, which have and / or include a conventional laptop and / or palmtop computer or other device with and / or including a radio frequency receiver. The "client", "terminal" and "terminal device" used herein may be portable, transportable, installed in a vehicle (air, sea and / or land), or suitable for and / or configured to run locally, and / or in a distributed form, at any other location on the earth and / or in space. The "client", "terminal" and "terminal device" used herein may also be a communication terminal, an Internet terminal, a music / video playing terminal, for example, a PDA, a MID (Mobile Internet Device) and / or a mobile phone with a music / video playing function, or a smart TV, a set-top box and other devices.

[0078] The hardware referred to by the names such as "server", "client", and "service node" in this application is essentially an electronic device with the equivalent capabilities of a personal computer. It is a hardware device with the necessary components revealed by the von Neumann principle, such as a central processing unit (including an arithmetic unit and a controller), a memory, an input device, and an output device. The computer program is stored in its memory, and the central processing unit calls the program stored in the external memory into the internal memory for execution, executes the instructions in the program, and interacts with the input and output devices to complete specific functions.

[0079] It should be pointed out that the concept of "server" referred to in this application can also be extended to the case of server clusters. According to the network deployment principle understood by those skilled in the art, the servers should be logically divided. In physical space, these servers can be independent of each other but can be called through interfaces, or integrated into a physical computer or a set of computer clusters. Those skilled in the art should understand this flexibility, and should not use it to restrict the implementation of the network deployment method of this application.

[0080] Unless expressly specified, one or more technical features of the present application can be deployed on a server for implementation and accessed by a client through a remote call to obtain an online service interface provided by the server, or can be directly deployed and run on a client for access.

[0081] The neural network models referenced or may be referenced in this application, unless expressly specified, can be deployed on a remote server and remotely called on the client, or can be deployed and directly called on a client with sufficient device capabilities. In some embodiments, when it runs on the client, its corresponding intelligence can be obtained through transfer learning to reduce the requirements for the client's hardware operating resources and avoid excessive occupation of the client's hardware operating resources.

[0082] Unless explicitly specified, the various data involved in this application can be stored remotely on a server or on a local terminal device, as long as it is suitable for being called by the technical solution of this application.

[0083] Those skilled in the art should be aware that, although the various methods of the present application are described based on the same concept and thus present commonality to each other, unless otherwise specified, these methods can be independently executed. Similarly, for each embodiment disclosed in the present application, they are all proposed based on the same inventive concept, therefore, concepts with the same expression, and concepts that are appropriately changed for convenience despite different expressions, should be understood as equivalent.

[0084] Unless the mutually exclusive relationship between the embodiments to be disclosed in this application is explicitly stated, the relevant technical features involved in each embodiment can be cross-combined to flexibly construct a new embodiment, as long as such combination does not deviate from the creative spirit of this application and can meet the needs of the prior art or solve certain deficiencies in the prior art. Those skilled in the art should be aware of this flexibility.

[0085] See also Figure 1 In one embodiment, the shopping mall store recommendation method of the present application includes:

[0086] Step S10, obtaining the number of store reviews and the store comprehensive score corresponding to each store in the shopping mall, determining the store popularity value corresponding to the store according to the number of store reviews and the store comprehensive score, determining the store popularity corresponding to the store according to the store popularity value, the store comprehensive score and the number of new reviews of the store within a preset time period, and determining the comprehensive recommendation value corresponding to each store according to the store popularity value and the store popularity;

[0087] The shopping mall store recommendation system can obtain the number of store reviews and the store comprehensive score corresponding to each store in the shopping mall, determine the store popularity value corresponding to the store according to the number of store reviews and the store comprehensive score, determine the store popularity corresponding to the store according to the store popularity value, the store comprehensive score and the number of new reviews of the store within a preset time period, and determine the comprehensive recommendation value corresponding to each store according to the store popularity value and the store popularity;

[0088] Specifically, in terms of the construction of multi-indicator data sets, the implementation process first involves the systematic collection of store information. The shopping mall store recommendation system uses public databases to obtain information such as the comprehensive score of each store, the number of reviews, and the number of new reviews in the past month to provide basic data for the recommendation system. These data not only provide support for subsequent recommendations, but also enable the system to conduct comprehensive data analysis. At the same time, the system obtains the user's consumption records and behavior data in the shopping mall through user authorization to ensure the accuracy and legality of the data. After obtaining this data, the system will conduct in-depth analysis to understand the user's preferences and habits, thereby laying the foundation for personalized recommendations.

[0089] Data preprocessing is a key step in ensuring data quality. The collected data needs to be cleaned, denoised and standardized to improve data quality and usability. Specifically, data cleaning includes removing duplicates, filling missing values ​​and processing outliers to avoid inaccurate data from having a negative impact on recommendation results. In addition, standardization is to unify data of different formats and sources so that subsequent algorithms can effectively process and ensure data integrity and consistency. To this end, the system uses advanced data processing technologies and tools to ensure data quality.

[0090] In some embodiments, the comprehensive score, number of reviews, number of new reviews in the past month and other information of each store are systematically collected from the public database and stored in the data warehouse; the collected data are cleaned to remove duplicates, erroneous information and missing values. After data cleaning, standardization is performed to ensure data consistency and reliability; based on the collected store information, the store heat value and store popularity value of each store are calculated: the heat value is processed by logarithmic scaling of the number of store reviews, and the calculation formula is:

[0091] Store popularity value = log (number of comments + 1),

[0092] The store popularity value can be generated by a weighted average model, combining the comprehensive score, the number of reviews, and the number of new reviews. The user's behavior data in the shopping mall, including the frequency of entering the store, purchase records, etc., can be collected through user authorization to ensure the legitimacy and privacy protection of the data. The multi-indicator data set is updated regularly (such as weekly or monthly) to ensure that the recommendation system can reflect the changes of users and stores in real time and maintain the accuracy and effectiveness of the system.

[0093] In a specific embodiment, the store popularity value corresponding to the store is determined according to the number of store reviews and the store comprehensive score. The calculation formula of the store popularity value is:

[0094]

[0095] Among them, Hotness represents the store popularity value corresponding to each store, R1 represents the number of store reviews received by the store, and r represents the comprehensive score of the store;

[0096] The store popularity corresponding to the store is determined according to the store popularity value, the store comprehensive score and the number of new reviews of the store within a preset time range. The calculation formula of the store popularity is:

[0097] P=α·Hotness+β·S+γ·ΔN,

[0098] Among them, P represents the store popularity corresponding to each store, S represents the comprehensive score of the store, ΔN represents the number of new reviews of the store within a preset time range, and the preset time range is within the past month, and α, β and γ represent weight coefficients;

[0099] The comprehensive recommendation value corresponding to each store is determined according to the store heat value and the store popularity. The calculation formula of the comprehensive recommendation value is:

[0100] R = λ·P+(1-λ)·Hotness,

[0101] Among them, R is the comprehensive recommendation value corresponding to each store, and λ is the weight coefficient, which is used to balance the importance of store popularity and store heat value.

[0102] Step S20, in response to an instruction to recommend a store to a target new user, obtaining a user interest category tag of the target new user and a store category tag corresponding to each store, calculating and determining a first similarity between the user interest category tag and the store category tag, determining a first recommendation degree corresponding to each store according to the comprehensive recommendation value, the first similarity, and an actual distance between the target new user and the store, and displaying each store in a descending order according to the first recommendation degree to a user interface of the target new user;

[0103] After determining the comprehensive recommendation value corresponding to each store according to the store heat value and the store popularity, the shopping mall store recommendation system can respond to the instruction to recommend stores to the target new user, obtain the user interest category label of the target new user and the store category label corresponding to each store, calculate and determine the first similarity between the user interest category label and the store category label, determine the first recommendation degree corresponding to each store according to the comprehensive recommendation value, the first similarity and the actual distance between the target new user and the store, and display each store from high to low according to the first recommendation degree to the user interface of the target new user;

[0104] Specifically, during the application of the recommendation algorithm, the system constructs a user-store matrix to match user preferences with the categories of stores in the shopping mall. For new users, the system provides a series of store category options when the user first enters the platform. These options will be used to generate user category labels for subsequent recommendation algorithms. The content-based recommendation algorithm combined with multiple indicators will then analyze the similarity between these labels and the store's category labels, and evaluate the recommendation by calculating the Jaccard similarity. This method not only effectively alleviates the cold start problem of new users, but also helps them quickly discover potential stores of interest.

[0105] In some embodiments, when a new user registers or logs in for the first time, basic information such as age, gender, shopping frequency, favorite product types, etc. is collected through a questionnaire or selection box, and this information is stored in a user database to facilitate subsequent construction of a user profile.

[0106] The shopping mall store recommendation system first performs a cluster analysis of all the stores in the shopping mall by subdivided categories. This step is to determine the type and number of clusters based on the subdivided categories of all the stores in the shopping mall, aiming to divide the stores into broader categories so that new users can choose more easily. After the clustering is completed, the system displays these store categories (for example: clothing, food, entertainment, home furnishings, etc.) in the interface after the user logs in, in the form of icons and text, so that users can choose the category of interest from them. The system records the user's choice and converts it into a category label to facilitate subsequent similarity calculations.

[0107] Furthermore, the user's category selection tags are generated, and the user's selected categories are mapped to a set of keywords, such as "clothing", "food", etc. These tags will be used in the subsequent similarity calculation and recommendation process.

[0108] In a specific embodiment, the first similarity between the user's interested category label and the store category label is calculated and determined, and the calculation formula of the first similarity is:

[0109]

[0110] Among them, Jaccard similarity represents the first similarity, A is the user's interested category label set, B is the store category label set, and the higher the Jaccard similarity, the greater the user's interest in the store;

[0111] The first recommendation degree corresponding to each store is determined according to the comprehensive recommendation value, the first similarity, and the actual distance between the target new user and the store. The calculation formula of the first recommendation degree is:

[0112]

[0113] Among them, ω1 represents the weight of the comprehensive recommendation value, ω2 represents the weight of the Jaccard similarity, ω3 represents the weight of the actual distance between the target new user and the store, and R represents the comprehensive recommendation value corresponding to each store.

[0114] In some embodiments, the step of determining the actual distance between the user and the store includes:

[0115] Step S201, obtaining a shopping center map to determine the starting location of a target new user or a target old user and the locations of multiple target stores;

[0116] Step S202: using an improved Dijkstra algorithm to perform path planning based on the starting point location and multiple target store locations to determine the actual distance from the current location of the target new user or the target old user to each target store.

[0117] Specifically, an improved Dijkstra algorithm is used to implement multi-objective path planning, which can quickly calculate and automatically display the actual walking distance of users in each store in the shopping mall. The system first automatically locates the user's current location, a process that does not require user intervention, ensuring the convenience of the user experience. Subsequently, the system imports detailed map data from the shopping mall's public database, including the precise coordinates of each store, areas where customers can move freely, and the locations of obstacles that may affect customers' walking. Using this data, combined with indoor positioning technology, the system can accurately determine the user's current location. Based on the user's current location, the improved Dijkstra algorithm is applied to calculate the shortest path from the user to each store.

[0118] The algorithm is specifically designed to handle multi-objective path planning problems, improving efficiency by maintaining a priority queue to ensure that the system always prioritizes the store closest to the user. When calculating the path, the algorithm takes into account the path length, estimated walking time, and the user's personal preferences to ensure that the most optimized walking route is provided. In addition, the system displays the calculated actual walking distance in real time next to each store entry, providing users with instant reference information. This real-time update and display of the real-time actual distance of each store relative to the user not only provides accurate actual walking distance, but also dynamically adjusts the recommendation list to ensure that users see the most relevant store options based on the actual distance indicator.

[0119] In the specific implementation process, the system first models the environment, builds a mobile model of the service object, and generates an obstacle grid map to simulate the actual layout of the shopping mall. In the path planning process, the system inputs the coordinates of the user's starting point and the target store into the algorithm to initialize the set of nodes to be visited and the set of nodes that have been visited. By looping through these nodes, the system continuously updates and optimizes the path until all target nodes are visited.

[0120] See also Figures 2 to 4 , the algorithm not only focuses on the shortest path when executing, but also considers the user's mobility model in real time, and updates the path in time to deal with sudden obstacles or path changes. For example, when the system detects a new obstacle on the user's path, it will immediately recalculate the optimal path to ensure that the user can reach the destination smoothly. This algorithm is designed to provide the best shopping experience and significantly reduce the time and inconvenience of users walking in the shopping mall.

[0121] In some embodiments, a floor plan of a shopping mall can be imported from a public database, including the coordinates of each store, the passable area, and the location information of obstacles. Ensure that the data format supports the input of the path planning algorithm. Set the user's starting point (current location) and the coordinates of the target store, call the improved Dijkstra algorithm, calculate the user's actual walking path to each target store, and record the shortest path and corresponding distance of each store for subsequent recommendation; during the user's walking process, monitor their current position in real time, obtain location data through sensors or mobile devices, and immediately re-plan the path if new obstacles or path changes are detected to ensure that the user can get the latest navigation information; display the calculated path to the user in a graphical manner through the mobile device interface, including walking routes, turn prompts and estimated arrival time, to enhance the user's shopping experience.

[0122] It can be seen from the above embodiments that the first recommendation degree corresponding to each store can be determined based on the comprehensive recommendation value corresponding to each store, the first similarity and the actual distance between the target new user and the store, and each store can be displayed to the user interface of the target new user from high to low according to the first recommendation degree.

[0123] Step S30, in response to the instruction to recommend a store to the old user, obtaining the user behavior time series data of the target old user, wherein the user behavior time series data includes historical consumption records and user consumption levels, the historical consumption records represent the store information visited and consumed by the user in different time periods, and the user consumption level is obtained by the user's consumption frequency in different stores and the price range of the stores;

[0124] The shopping mall store recommendation system can respond to the instruction of recommending stores to old users and obtain the user behavior time series data of the target old users, wherein the user behavior time series data includes historical consumption records and user consumption levels, the historical consumption records represent the store information visited and consumed by users in different time periods, and the user consumption level is obtained by the consumption frequency of users in different stores and the price range of the stores;

[0125] Specifically, for old users, the shopping mall store recommendation system adopts a collaborative filtering recommendation algorithm that integrates the long short-term memory network model (LSTM model). First of all, the LSTM model has a strong ability in time series data processing and can capture the dynamic changes of user interests. The system will input the user's behavior data in time series to train the user's interest in some types of stores in the current time period. The input user behavior data includes the user's consumption in a certain type of store at a certain point in time. In model training, we also used mean square error as the loss function, and by using the Adam optimizer, which combines the advantages of momentum and adaptive learning rate, it helps the model converge quickly and improve accuracy, effectively avoiding overfitting.

[0126] Secondly, the system uses the quantitative results obtained by the LSTM model to construct a user-store matrix. In order to optimize the personalized recommendation effect, we also take the user's consumption level into consideration and add the consumption level to the construction of the user-store matrix. This indicator is calculated based on the user's consumption frequency in different stores and the store's price range, thus providing the system with a method to quantify the user's economic ability. The calculation formula for the user's consumption level is:

[0127]

[0128] Among them, T i represents the number of times or frequency of consumption by the user in the i-th store, P i Indicates the price range of the store, and X represents the user's consumption level. In this way, the user's consumption level can be used as a weight factor in the recommendation algorithm to ensure that the recommendation results are both personalized and highly relevant. By incorporating the consumption level as a weight factor into the recommendation algorithm, the recommendation results are both personalized and highly relevant.

[0129] Step S40: using a long short-term memory network model trained with the user behavior time series data to predict the target old user's stores of interest in the current period, and using a preset collaborative filtering algorithm to calculate and determine the second similarity between the target old user and other users based on the stores of interest to the target old user in the current period, so as to determine the target old user's neighbor users;

[0130] After obtaining the user behavior time series data of the target old user, a long short-term memory network model trained with the user behavior time series data is used to predict the target old user's stores of interest in the current period, and a preset collaborative filtering algorithm is used to calculate and determine the second similarity between the target old user and other users based on the stores of interest to the target old user in the current period, so as to determine the neighbor users of the target old user;

[0131] Specifically, detailed information of old users can be collected, including historical consumption records, visit frequency, preferred store types, etc. These data can be obtained through user behavior analysis and database query. Model training is performed based on user behavior time series data. First, the user's consumption records of stores are extracted as training data sets. These records include information about stores visited and consumed by users in different time periods. Using the LSTM model, these time series data can be trained to obtain quantitative values ​​of the target old users' interest in various stores, thereby effectively capturing the dynamic changes in user interests. After the model training is completed, a user-store matrix is ​​constructed based on the quantitative results obtained from the model training. The rows of the matrix represent users, the columns represent stores, and each element of the matrix represents the user's rating of the corresponding store.

[0132] To further optimize the recommendation effect, the system also calculates the user's consumption level, which is a comprehensive indicator based on the user's consumption frequency in different stores and the store's price range. Finally, the user's consumption level is incorporated into the construction of the user-store matrix to achieve more accurate personalized recommendations.

[0133] In a further embodiment, the historical consumption records and the user's consumption level are used as a training set, and the long short-term memory network model (LSTM model) trained by the training set is used to obtain the quantitative value of the target old user's interest in various stores, thereby effectively capturing the dynamic changes of user interests. After the model training is completed, a user-store matrix is ​​constructed based on the quantitative results obtained by the model training, where the rows of the matrix represent users, the columns represent stores, and each element of the matrix represents the user's rating of the corresponding store, so as to predict the target old user's store of interest in the current period.

[0134] As shown in Table 1, Table 1 shows the user-store matrix of five users, which includes the users' interest in different store categories and their consumption levels. Through this matrix, the system can understand each user's preference intensity for each type of store and the user's own consumption level, so as to predict the stores that the target old users are interested in during the current period.

[0135] Table 1 User-store matrix of five users

[0136]

[0137] In a further embodiment, as shown in Table 2, in order to further enhance the relevance and accuracy of the recommendation, the system also generates a user-user similarity matrix to discover potential similar interest groups by comparing the preference similarities between different users. This similarity matrix is ​​constructed based on the user's behavioral preferences on store categories, which helps the system understand which users may have similar shopping interests, thereby providing more personalized recommendations for each user.

[0138] Table 2 Similarity matrix between five users

[0139]

[0140] Table 2 shows the similarity matrix between the five users, where the closer the value is to 1, the more similar the preferences between the users are. By analyzing this matrix, the system can identify user groups with similar shopping preferences and provide them with more targeted recommendations.

[0141] In some embodiments, a preset collaborative filtering algorithm is used to calculate and determine the second similarity between the target old user and other users according to the stores that the target old user is interested in during the current period. The calculation formula of the second similarity is expressed as:

[0142]

[0143] Among them, sim(u,v) represents the second similarity, u represents the vector of the target old user, v represents the vector of other users, u and v contain information about the user's interest in the store and consumption level, θ represents the angle between user vectors, and the value range of cosine similarity cos(θ) is [0,1]. The larger the value, the more similar other users are to the target old user.

[0144] The second similarity between the target old user and other users can be calculated and determined by the above calculation formula to determine the neighbor users of the target old user.

[0145] Step S50: predict the estimated ratings of the target old user for the unrated stores based on the ratings of the neighbor users for each store, calculate and determine the second recommendation degree corresponding to each store based on the estimated ratings, the actual distance between the target old user and the store, and the comprehensive recommendation value, and display each store to the user interface of the target old user from high to low according to the second recommendation degree to complete the store recommendation of the shopping mall.

[0146] A preset collaborative filtering algorithm is used to calculate the second similarity between the target old user and other users based on the stores that the target old user is interested in within the current time period. After determining the neighbor users of the target old user, the estimated rating value of the target old user for the stores that have not been rated is predicted based on the rating values ​​of the neighbor users for each store. The second recommendation degree corresponding to each store is calculated and determined based on the estimated rating value, the actual distance between the target old user and the store, and the comprehensive recommendation value. Each store is displayed to the user interface of the target old user from high to low according to the second recommendation degree, so as to complete the store recommendation of the shopping mall.

[0147] Specifically, the calculation of the second recommendation degree corresponding to each store is a key link to ensure the accuracy and effectiveness of the recommendation results. The system will set appropriate weights for each indicator based on the user's interests, the actual distance of the store, and the comprehensive recommendation value. Through weighted addition, the final second recommendation degree corresponding to each store is generated, guiding the system to show users the recommendation results that best meet their needs. This process not only considers the user's immediate needs, but also takes into account long-term preferences, ensuring that the recommendation system can optimize the user experience in a constantly changing environment.

[0148] Throughout the implementation process, the system continuously collects user feedback and behavior data to adjust the recommendation algorithm and path planning strategy in real time. By continuously updating the user portrait and store feature model, the system can better adapt to the changing needs of users. In addition, the system regularly conducts data analysis and performance evaluation to ensure the effectiveness and accuracy of the recommendation results, thereby continuously improving user satisfaction.

[0149] In a specific embodiment, the estimated rating of the target old user for the unrated store is predicted based on the rating of each store by the neighbor user. The calculation formula of the estimated rating is:

[0150]

[0151] in, is the estimated rating value, which represents the predicted rating of the target old user u for the store i, N represents the set of neighbor users with the highest similarity to the target old user u, ω uv represents the similarity between the target old user u and the neighbor user v, r vi represents the rating of store i by neighbor user v;

[0152] According to the estimated rating value, the actual distance between the target old user and the store, and the comprehensive recommendation value, the second recommendation corresponding to each store is calculated and determined, and the calculation formula of the second recommendation is:

[0153]

[0154] Among them, α, β and γ represent weight coefficients, and R represents the comprehensive recommendation value corresponding to each store.

[0155] Through the above implementation steps, this application realizes efficient path planning and personalized recommendations, and provides a flexible and intelligent shopping experience platform. The system can respond to user needs in real time in a rapidly changing shopping environment, thereby significantly improving user satisfaction and providing a feasible technical foundation for the development of future smart shopping centers.

[0156] In some embodiments, with the rapid development of the current Internet and mobile technology, the number of stores in shopping malls continues to increase, and users often feel lost when faced with a large number of choices. Although traditional recommendation systems can make personalized recommendations by analyzing users' historical behavior data, they still face many challenges, including the lack of sufficient historical interaction data for new users and the rapid changes in user preferences. Old users face difficulties such as the incompleteness of data due to the increasing number of users and types of items, which makes the interaction between users and items more and more sparse. Therefore, it is particularly important to develop an efficient and accurate recommendation method.

[0157] In this application, in order to predict the user's interest in shopping mall stores, we designed three prediction models based on different recursive neural network architectures: RNN model, LSTM model, and GRU model. Through comparative experiments, we evaluated the performance of these three models on key performance indicators, including training time, inference time, training loss, validation loss, and accuracy.

[0158] See also Figure 5 and Figure 6 , Figure 5 and Figure 6 The comparison of training time and inference time of RNN model, LSTM model and GRU model is shown respectively. The results show that although the training time of LSTM model is slightly longer than RNN, its inference time is similar to that of RNN and both are better than GRU model. This finding is crucial for real-time recommendation systems that need to respond quickly to user requests, because LSTM model can meet the high real-time requirements of the system while ensuring prediction accuracy.

[0159] See also Figure 7 , Figure 8 and Fig. 9 , Figure 7 , Figure 8 and Fig. 9The performance comparison of the three models in terms of training loss, validation loss and accuracy is revealed respectively. It can be observed from the curves of training loss and validation loss that the loss of the LSTM model decreases more steadily during the training process and reaches a lower loss value in the same training cycle. In addition, the accuracy of the LSTM model is also significantly higher than that of the RNN and GRU models, which shows that the LSTM model has significant advantages in capturing long-term dependencies and processing time series data. Taking into account the prediction accuracy, training efficiency and inference efficiency of the model, the LSTM model was selected as the key component in the recommendation system of this application due to its excellent performance in various indicators. The introduction of the LSTM model not only improves the accuracy of user interest prediction, but also enhances the stability and reliability of the recommendation system when processing complex user behavior data.

[0160] In traditional recommendation systems, distance metrics are usually estimated using Manhattan distance or Euclidean distance. However, these methods cannot accurately reflect the path and distance that users actually walk. To solve this problem, this application proposes an improved Dijkstra algorithm to accurately calculate the actual walking distance of users to each store. The algorithm updates the path in real time by cyclic backtracking to provide users with the optimal walking route. This innovation not only improves the accuracy of path planning, but also provides an important actual distance indicator for the recommendation system.

[0161] In the recommendation strategy for new users, considering that their main problem is the lack of historical data, the recommendation system needs to comprehensively analyze the user-selected tags and the characteristics of popular stores in the current shopping mall. We innovatively adopt a content-based recommendation algorithm that combines multiple indicators to provide diversified recommendations. Specifically, by analyzing the Jaccard similarity coefficient between the category tags selected by the user and the store category tags, the system can effectively alleviate the cold start problem of new users and help them quickly discover stores of potential interest. In addition, combined with the comprehensive recommendation value of the store, the similarity coefficient, and the actual distance between the user and the store, a weighted formula is used to calculate the final recommendation degree to ensure that the recommendation results meet user needs to the greatest extent.

[0162] For old users, the system uses a collaborative filtering recommendation algorithm that integrates neural networks. The user's historical behavior data is trained through the LSTM model, and the obtained quantitative results form a user-item matrix and calculate the similarity between users. Introducing the user's consumption level into the user-item matrix as a weight factor can more realistically reflect the user's preference for each store. In order to cope with the sparsity problem of user-item interaction, this application adopts a clustering method of subdivided categories. Specifically, by clustering the stores, the store types are converted into unique hot encodings, thereby improving the processability of the data. The clustering process is based on the subdivided categories of the stores, and a large number of store types are summarized into several major categories, each of which is represented by a unique label. This step not only simplifies the data structure, but also facilitates the subsequent prediction of user interests, improves the density of the data, and then uses LSTM for training. This strategy effectively captures the user's interest characteristics and enhances the accuracy of the recommendation.

[0163] In addition, the system also adopts a user similarity score prediction method to more comprehensively predict the interest of old users in stores. When old users do not have direct ratings or interaction records for certain store types, the system will look for other user groups with high similarity to these users and analyze the ratings and feedback of these neighboring users on the target store. By calculating the similarity weights between the target user and the neighboring users, the system can reasonably predict the target user's potential ratings for unrated stores. This prediction method based on neighboring user ratings can not only supplement the missing data in the user-item matrix, but also improve the accuracy and diversity of recommendation results. When implementing this method, the system first identifies neighboring user groups with similar consumption habits and preferences as the target user, then collects the ratings and feedback of these neighboring users on the target store, and then performs a weighted average calculation based on the similarity between the target user and the neighboring users to obtain the target user's predicted rating for the store, and applies the predicted rating to the recommendation algorithm to dynamically update the target user's recommendation list to ensure the relevance and novelty of the recommended content. By combining the user-item matrix and user similarity score prediction, the system can provide richer and more personalized store recommendations for old users, and ensure the accuracy and timeliness of the recommendation results even when the user-item interaction data is sparse.

[0164] In the core part of the recommendation algorithm, this application focuses on comprehensively considering user behavior data, store characteristics and real-time data to generate accurate personalized recommendations. The system builds a multi-dimensional feature model of the store by establishing a comprehensive multi-indicator data set to collect the comprehensive score, number of comments and number of new comments of each store in the past month. During the data processing process, the logarithmic scaling method is used to process the long-tail distribution of the number of comments, and the score and number of comments are comprehensively considered to calculate the popularity value of each store. In addition, the popularity value is calculated using a weighted average model, which comprehensively considers the comprehensive score, number of comments and number of new comments, and sets different weights according to the importance of each indicator, and finally generates a comprehensive recommendation value for each store, providing a solid foundation for subsequent recommendation algorithms.

[0165] In terms of user data processing, the system records the price range of users entering the store and quantifies it into values ​​from 1 to 4 to represent the price levels from low to high. By counting the frequency of users visiting each store, combined with the price range and the number of visits, the system can more accurately estimate the user's consumption level, generate user portraits, and lay the foundation for personalized recommendations. In addition, the system also constructs a user-user matrix, and further enhances the performance of the recommendation system by calculating the similarity between different users. This similarity calculation can help the system better understand the preference distribution of user groups, thereby providing more personalized recommendations for each user.

[0166] This application verifies the effectiveness of the proposed method through experiments, including the training and evaluation of the neural network model, the verification of the recommendation algorithm, and the testing of the path planning algorithm. The experimental results show that the LSTM model can better capture user interests, and the recommendation algorithm can effectively discover the user's hidden preferences, verifying the superiority of the recommendation system of this application in terms of accuracy, real-time and user satisfaction.

[0167] It can be seen from the above embodiments that, compared with the prior art, the present application aims at the following problems: the interaction between new users and items in the prior art is very sparse, which makes it more difficult to capture the real preferences of users; for old users, as the number of users and the types of items increase, the interaction between users and items becomes increasingly sparse, making it difficult for the system to accurately capture the real preferences of users. The present application includes but is not limited to the following beneficial effects:

[0168] First, through the improved Dijkstra algorithm, the system can accurately calculate the actual walking distance of users to each store, taking into account the actual walking path of users, rather than just the traditional distance measurement (such as Manhattan distance or Euclidean distance). This precise path planning greatly improves the user's navigation experience in the shopping mall, provides users with the best walking route, and further optimizes the shopping process;

[0169] Secondly, this application solves the cold start problem for new users. In the new user recommendation strategy, by combining the user-selected tags with the characteristics of popular stores in the current shopping mall, the system can effectively alleviate the cold start problem. The Jaccard similarity coefficient is used to analyze the relationship between the category tags selected by the user and the store category tags, so that new users can quickly find stores that meet their potential interests.

[0170] Third, for old users, the system uses an LSTM neural network model to analyze the user's historical behavior data and accurately capture the user's interest characteristics. Combined with the clustering method of subdivided categories, the sparsity problem is reduced, and the accuracy of recommendations and the processability of data are improved. By weighted calculation of the weight factors in the user-item matrix, the system can more realistically reflect the user's preferences for each store and improve the personalization of recommendations.

[0171] Fourthly, this application uses the user similarity score prediction method to enable the system to more comprehensively predict the user's interest in a store. When a user lacks direct interaction records with certain stores, the system can predict the target user's potential preferences through the scores and feedback of neighboring users. This weighted prediction method based on neighbor users effectively fills the gaps in the user-item matrix and further improves the accuracy and diversity of recommendations.

[0172] Fifth, at the core of the recommendation algorithm, the system comprehensively considers data from multiple dimensions (such as store ratings, number of reviews, new reviews, etc.), and calculates the popularity and heat value of the store through a weighted average model. This multi-dimensional comprehensive calculation ensures the accuracy and timeliness of the recommendation results, making the recommended content more in line with the interests and needs of users.

[0173] Sixth, by collecting statistics on the price range and visit frequency of users entering the store, the system can more accurately build user portraits, analyze user consumption levels, and provide a more reliable basis for personalized recommendations. The similarity calculation between users further enhances the system's recommendation capabilities, thereby ensuring that each user can receive the recommendations that best meet their needs.

[0174] Seventh, this application improves user satisfaction and shopping experience. Combined with accurate route planning, real-time updates and personalized recommendations, the system can greatly improve the user's shopping experience, reduce unnecessary time waste, and help users quickly find stores that match their interests. Overall, the recommendation system can significantly improve user satisfaction and increase the flow and user stickiness of shopping malls.

[0175] In summary, this application implements a recommendation system that comprehensively utilizes user behavior data, store characteristics, and actual distance. Through the combination of deep learning models and multiple algorithms, it provides users with personalized store recommendations, significantly improving user experience and customer satisfaction. This technical solution not only has good practicality, but also contributes to the construction of future smart shopping navigation systems, adding more convenience and fun to users' shopping experience. Combining positioning technology, real-time data analysis, and personalized recommendation algorithms, this application shows broad application prospects and can provide strong support for future smart retail and shopping center management.

[0176] See also Fig.10A shopping mall store recommendation device provided to meet one of the purposes of the present application includes a comprehensive recommendation value determination module 1100, a new user store recommendation module 1200, a user behavior data acquisition module 1300, a neighbor user determination module 1400 and an old user store recommendation module 1500. Among them, the comprehensive recommendation value determination module 1100 is configured to obtain the number of store reviews and the store comprehensive score corresponding to each store in the shopping mall, determine the store popularity value corresponding to the store according to the number of store reviews and the store comprehensive score, determine the store popularity corresponding to the store according to the store popularity value, the store comprehensive score and the number of new reviews of the store within a preset time range, and determine the comprehensive recommendation value corresponding to each store according to the store popularity value and the store popularity; the new user store recommendation module 1200 is configured to respond to the instruction of recommending stores to the target new user, obtain the user interest category label of the target new user and the store category label corresponding to each store, calculate and determine the first similarity between the user interest category label and the store category label, determine the first recommendation degree corresponding to each store according to the comprehensive recommendation value, the first similarity and the actual distance between the target new user and the store, and display each store from high to low according to the first recommendation degree to the user interface of the target new user; the user behavior data acquisition module 1300 is configured to respond to the instruction of recommending stores to the old user, obtain the user interest category label of the target old user, and the store category label corresponding to each store. The user behavior time series data includes historical consumption records and user consumption levels, wherein the historical consumption records represent information about stores visited and consumed by users in different time periods, and the user consumption levels are obtained by the consumption frequencies of users in different stores and the price ranges of the stores; a neighbor user determination module 1400 is configured to use a long short-term memory network model trained with the user behavior time series data to predict the stores of interest to the target old user in the current time period, and to use a preset collaborative filtering algorithm to calculate and determine the second similarity between the target old user and other users based on the stores of interest to the target old user in the current time period, so as to determine the neighbor users of the target old user; an old user store recommendation module 1500 is configured to predict the estimated rating value of the target old user for the unrated stores according to the rating values ​​of the neighbor users for each store, and to calculate and determine the second recommendation degree corresponding to each store according to the estimated rating value, the actual distance between the target old user and the store, and the comprehensive recommendation value, and to display each store from high to low according to the second recommendation degree to the user interface of the target old user, so as to complete the store recommendation of the shopping mall.

[0177] Based on any embodiment of this application, please refer to Fig.11 Another embodiment of the present application further provides an electronic device, which can be implemented by a computer device, such as Fig.11 As shown, a schematic diagram of the internal structure of a computer device. The computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. Among them, the computer-readable storage medium of the computer device stores an operating system, a database, and computer-readable instructions. The database may store a control information sequence. When the computer-readable instructions are executed by the processor, the processor may implement a shopping mall store recommendation method. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. The memory of the computer device may store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor may execute the shopping mall store recommendation method of the present application. The network interface of the computer device is used to connect and communicate with a terminal. Those skilled in the art will understand that Fig.11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0178] In this embodiment, the processor is used to execute Fig.10 The memory stores the program code and various data required to execute the above modules or submodules. The network interface is used to transmit data between user terminals or servers. The memory in this embodiment stores the program code and data required to execute all modules / submodules in the shopping mall store recommendation device of this application, and the server can call the program code and data of the server to execute the functions of all submodules.

[0179] The present application also provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the shopping mall store recommendation method described in any embodiment of the present application.

[0180] The present application also provides a computer program product, including a computer program / instruction, which, when executed by one or more processors, implements the steps of the shopping mall store recommendation method described in any embodiment of the present application.

[0181] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments of the present application can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a computer-readable storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0182] The above description is only a partial implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

[0183] In summary, this application implements a recommendation system that comprehensively utilizes user behavior data, store characteristics, and actual distance. Through the combination of deep learning models and multiple algorithms, it provides users with personalized store recommendations, significantly improving user experience and customer satisfaction. This technical solution not only has good practicality, but also contributes to the construction of future smart shopping navigation systems, adding more convenience and fun to users' shopping experience. Combining positioning technology, real-time data analysis, and personalized recommendation algorithms, this application shows broad application prospects and can provide strong support for future smart retail and shopping center management.

Claims

1. A method for recommending stores in a shopping mall, characterized in that: include: Obtain the number of store reviews and the store comprehensive score corresponding to each store in the shopping mall, determine the store popularity value corresponding to the store according to the number of store reviews and the store comprehensive score, determine the store popularity corresponding to the store according to the store popularity value, the store comprehensive score and the number of new reviews of the store within a preset time period, and determine the comprehensive recommendation value corresponding to each store according to the store popularity value and the store popularity; In response to an instruction to recommend a store to a target new user, obtain a user interest category tag of the target new user and a store category tag corresponding to each store, calculate and determine a first similarity between the user interest category tag and the store category tag, determine a first recommendation degree corresponding to each store according to the comprehensive recommendation value, the first similarity, and an actual distance between the target new user and the store, and display each store in a descending order according to the first recommendation degree to a user interface of the target new user; In response to an instruction to recommend a store to an old user, the user behavior time series data of the target old user is obtained, wherein the user behavior time series data includes historical consumption records and user consumption levels, the historical consumption records represent information about stores visited and consumed by the user in different time periods, and the user consumption level is obtained by the user's consumption frequency in different stores and the price range of the stores; A long short-term memory network model trained with the user behavior time series data is used to predict the target old user's stores of interest in the current period, and a preset collaborative filtering algorithm is used to calculate and determine the second similarity between the target old user and other users based on the stores of interest to the target old user in the current period, so as to determine the neighbor users of the target old user; The estimated ratings of the target old user for the unrated stores are predicted based on the ratings of the neighbor users for each store. The second recommendation degree corresponding to each store is calculated and determined based on the estimated ratings, the actual distance between the target old user and the store, and the comprehensive recommendation value. Each store is displayed to the user interface of the target old user from high to low according to the second recommendation degree, so as to complete the store recommendation of the shopping mall.

2. The method for recommending shopping mall stores according to claim 1, characterized in that: The store popularity value corresponding to the store is determined according to the number of store reviews and the store comprehensive score. The calculation formula of the store popularity value is: Among them, Hotness represents the store popularity value corresponding to each store, R1 represents the number of store reviews received by the store, and r represents the comprehensive score of the store; The store popularity corresponding to the store is determined according to the store popularity value, the store comprehensive score and the number of new reviews of the store within a preset time range. The calculation formula of the store popularity is: P=α·Hotness+β·S+γ·ΔN, Among them, P represents the store popularity corresponding to each store, S represents the comprehensive score of the store, ΔN represents the number of new reviews of the store within a preset time range, and the preset time range is within the past month, and α, β and γ represent weight coefficients; The comprehensive recommendation value corresponding to each store is determined according to the store heat value and the store popularity. The calculation formula of the comprehensive recommendation value is: R = λ·P+(1-λ)·Hotness, Among them, R is the comprehensive recommendation value corresponding to each store, and λ is the weight coefficient, which is used to balance the importance of store popularity and store heat value.

3. The method for recommending shopping mall stores according to claim 1, characterized in that: The first similarity between the user's interested category label and the store category label is calculated and determined, and the calculation formula of the first similarity is: Among them, Jaccard similarity represents the first similarity, A is the user's interested category label set, B is the store category label set, and the higher the Jaccard similarity, the greater the user's interest in the store; The first recommendation degree corresponding to each store is determined according to the comprehensive recommendation value, the first similarity, and the actual distance between the target new user and the store. The calculation formula of the first recommendation degree is: Among them, ω1 represents the weight of the comprehensive recommendation value, ω2 represents the weight of the Jaccard similarity, ω3 represents the weight of the actual distance between the target new user and the store, and R represents the comprehensive recommendation value corresponding to each store.

4. The method for recommending shopping mall stores according to claim 1, characterized in that: The calculation formula of the user consumption level is: Among them, T i represents the number of times or frequency of consumption by the user in the i-th store, P i represents the price range of the store, and X represents the user's consumption level.

5. The method for recommending shopping mall stores according to claim 1, characterized in that: The preset collaborative filtering algorithm is used to calculate and determine the second similarity between the target old user and other users according to the stores that the target old user is interested in during the current period. The calculation formula of the second similarity is expressed as: Among them, sim(u,v) represents the second similarity, u represents the vector of the target old user, v represents the vector of other users, u and v contain information about the user's interest in the store and consumption level, θ represents the angle between user vectors, and the value range of cosine similarity cos(θ) is [0,1]. The larger the value, the more similar other users are to the target old user.

6. The method for recommending shopping mall stores according to claim 1, characterized in that: The estimated ratings of the target old users for the unrated stores are predicted based on the ratings of the neighbor users for each store. The calculation formula for the estimated ratings is: in, is the estimated rating value, which represents the predicted rating of the target old user u for the store i, N represents the set of neighbor users with the highest similarity to the target old user u, ω uv represents the similarity between the target old user u and the neighbor user v, r vi represents the rating of store i by neighbor user v; According to the estimated rating value, the actual distance between the target old user and the store, and the comprehensive recommendation value, the second recommendation corresponding to each store is calculated and determined, and the calculation formula of the second recommendation is: Among them, α, β and γ represent weight coefficients, and R represents the comprehensive recommendation value corresponding to each store.

7. The method for recommending shopping mall stores according to claim 1, characterized in that: The steps for determining the actual distance between the user and the store include: Obtain a shopping mall map to determine the starting location of a target new user or target old user and multiple target store locations; An improved Dijkstra algorithm is used to perform path planning based on the starting point location and multiple target store locations to determine the actual distance from the current location of the target new user or the target old user to each target store.

8. A shopping mall store recommendation device, characterized in that: include: A comprehensive recommendation value determination module is configured to obtain the number of store reviews and the store comprehensive score corresponding to each store in the shopping mall, determine the store popularity value corresponding to the store according to the number of store reviews and the store comprehensive score, determine the store popularity corresponding to the store according to the store popularity value, the store comprehensive score and the number of new reviews of the store within a preset time period, and determine the comprehensive recommendation value corresponding to each store according to the store popularity value and the store popularity; A new user store recommendation module is configured to respond to an instruction to recommend a store to a target new user, obtain a user interest category tag of the target new user and a store category tag corresponding to each store, calculate and determine a first similarity between the user interest category tag and the store category tag, determine a first recommendation degree corresponding to each store according to the comprehensive recommendation value, the first similarity, and an actual distance between the target new user and the store, and display each store in a descending order according to the first recommendation degree to a user interface of the target new user; A user behavior data acquisition module, configured to respond to an instruction to recommend a store to an old user, and acquire user behavior time series data of a target old user, wherein the user behavior time series data includes historical consumption records and user consumption levels, the historical consumption records represent information about stores visited and consumed by users in different time periods, and the user consumption level is obtained by the consumption frequency of users in different stores and the price range of the stores; A neighbor user determination module is configured to use a long short-term memory network model trained with the user behavior time series data to predict the target old user's stores of interest in the current period, and use a preset collaborative filtering algorithm to calculate and determine the second similarity between the target old user and other users based on the stores of interest to the target old user in the current period, so as to determine the neighbor users of the target old user; The old user store recommendation module is configured to predict the estimated rating value of the target old user for the unrated stores based on the rating values ​​of the neighbor users for each store, calculate and determine the second recommendation degree corresponding to each store based on the estimated rating value, the actual distance between the target old user and the store, and the comprehensive recommendation value, and display each store to the user interface of the target old user from high to low according to the second recommendation degree, so as to complete the store recommendation of the shopping mall.

9. An electronic device, comprising a central processing unit and a memory, characterized in that: The central processing unit is used to call and run the computer program stored in the memory to execute the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: It stores a computer program implemented according to the method described in any one of claims 1 to 7 in the form of computer-readable instructions, and when the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

Images