Multi-store O2O smart mall system based on Ai

Abstract

The invention relates to the technical field of e-commerce, and discloses an Ai-based multi-store O2O smart mall system, which comprises a user portrait module, a store portrait module and a recommendation module. According to the application, the user portrait module integrates the online data to generate the first user portrait, and updates the first user portrait to the second user portrait in combination with the offline data, thereby comprehensively capturing user demand changes; the store portrait module is fused with online and offline data to construct a store portrait, and the operation quality and the supply capability are accurately reflected; the recommendation module dynamically screens and recommends stores based on portraits, updates the stores in real time, and generates a scenarized scheme in combination with multi-target optimization; precise matching between the user and the store is achieved, recommendation accuracy and timeliness are improved, user offline experience is optimized, the purchase conversion rate is improved, meanwhile, the store is assisted to efficiently allocate resources, and the operation efficiency and commercial value of a multi-store O2O mode are enhanced.

Description

Technical Field

[0001] This application relates to the field of e-commerce technology, specifically an AI-based multi-store O2O smart mall system. Background Technology

[0002] In today's e-commerce field, the O2O (online-to-offline) model has become a mainstream trend, and the demand for multi-store O2O smart mall systems is growing. However, existing related technologies still have significant shortcomings.

[0003] Traditional e-commerce recommendation systems primarily generate user profiles based on individual users' online behavior data, such as browsing history and purchase history, but neglect users' actual behavior and experience data in offline stores. This single-dimensional profile cannot comprehensively and accurately reflect users' true needs and changes in preferences, leading to significant discrepancies between recommendation results and users' actual purchasing behavior.

[0004] Meanwhile, existing systems typically treat stores as a uniform source of goods, failing to adequately consider the differences in operational quality and supply capacity among different stores, such as inventory levels, delivery timeliness, and service quality. This can lead to recommended stores failing to meet users' personalized needs; for example, recommended stores might be too far from the user, have insufficient inventory, or offer poor service, thus impacting user experience and purchase conversion rates.

[0005] Furthermore, existing multi-store O2O systems lack an effective real-time update mechanism, making it impossible to dynamically adjust recommendation results based on user and store offline activities. User behavior at physical stores, such as trying on clothes, using products, and consulting, as well as real-time changes in store operations, such as promotional activities and inventory updates, are not promptly reflected in the recommendation system. This results in recommendations lagging behind actual scenarios, reducing the system's usability and commercial value.

[0006] Chinese invention patent application CN113888281A discloses a product recommendation method, device, system and storage medium, but the invention performs poorly in handling differences at the store level.

[0007] In conclusion, there is an urgent need for a new AI-based multi-store O2O smart mall technical solution. Summary of the Invention

[0008] The purpose of this application is to provide an AI-based multi-store O2O smart mall system to solve the technical problems mentioned in the background.

[0009] To achieve the above objectives, this application discloses the following technical solution: a multi-store O2O smart mall system based on AI, the system comprising:

[0010] The user profile module is used to obtain the first user profile through online user data. The first user profile is used to define the user's online intention for products.

[0011] The store profile module is used to obtain the first store profile through online store data. The first store profile is used to define the store's operating quality and supply capacity.

[0012] The recommendation module is used to filter the first store profile based on the first user profile to obtain at least one recommended store that responds to the first user profile; wherein the recommended stores are updated in real time based on the offline activities of users and stores.

[0013] Preferably, obtaining the first user profile from online user data includes the following steps:

[0014] The user's online data is acquired, including the user's online browsing history, search keywords, favorited products, add-to-cart behavior, historical order data, and review data.

[0015] Feature extraction and cluster analysis are performed on the user's online data to obtain the corresponding preference vector, which includes product category preference, price sensitivity, brand preference, and purchase frequency characteristics.

[0016] The first user profile is obtained based on the preference vector. The first user profile includes the user's interest rating for different categories of goods and prediction of potential demand.

[0017] Preferably, obtaining the first store profile through online store data includes the following steps:

[0018] The online data of the stores is obtained, including the stores' product inventory data, historical order fulfillment rate, average delivery time, user review data, promotional activity records, and regional coverage.

[0019] The online data of the stores is standardized to obtain a set of online indicators for the stores, which includes indicators of the store's product richness, service quality, and fulfillment capability.

[0020] The first store profile is obtained based on the online indicator set of the store. The first store profile includes the store's operational quality scoring system and commodity supply capacity feature matrix.

[0021] Preferably, the step of filtering the first store profile based on the first user profile includes the following steps:

[0022] The similarity score is calculated by comparing the first user profile with the product supply capacity feature matrix. The similarity score is used to define the store's corresponding capacity for the first user profile.

[0023] The stores are ranked by weight based on the aforementioned operational quality scoring system;

[0024] Stores that do not meet the requirements are filtered out by constraints, including user-specified delivery range, expected delivery time, and specific product inventory thresholds.

[0025] Preferably, obtaining at least one recommended store in response to the first user profile includes the following steps:

[0026] Output the stores whose similarity scores are greater than a preset similarity score threshold and whose business quality scores are in the top N, and obtain a store list. Then, determine the stores in the store list as the recommended stores.

[0027] For each store in the store list, a personalized recommended product combination is generated. The product combination is generated based on the store's inventory and the user's interest rating for each product category. The personalized recommended product combination is pushed to the user's terminal, and the personalized recommended product combination and its corresponding user information are pushed to the recommended store.

[0028] Preferably, when a user visits different recommended stores, the user profiling module obtains a second user profile through the user's offline data. The second user profile is used to define the changes in the user's intention to purchase products in the first user profile after the offline activity.

[0029] The store profile module is used to obtain a second store profile through offline store data. The second store profile is used to define the response of the recommended store to the user's offline activities.

[0030] The recommendation module is used to filter the second store profile based on the second user profile to obtain a sorted list of recommended stores in response to the second user profile.

[0031] Preferably, obtaining the second user profile through offline user data includes the following steps:

[0032] The system acquires the user's offline data generated at the recommended stores, including browsing history, trial records, offline transaction data, store service evaluations, and membership points usage records.

[0033] Spatiotemporal features are extracted from the user's offline data to construct the user's offline behavior sequence, which includes the user's dwell time in each store, product interaction depth, and decision path.

[0034] By comparing and analyzing the user's offline data with the first user profile, the change rate of user interest in different product categories is obtained. The change rate of interest includes a positive increase factor and a negative decrease factor.

[0035] The potential demand prediction is updated based on the interest change rate, and a second user profile is generated. The second user profile includes a preference correction coefficient matrix for users in offline scenarios.

[0036] Preferably, obtaining the second store profile through offline store data includes the following steps:

[0037] Obtain the offline data of each store, including real-time customer flow data, merchandise display adjustment records, offline promotional activity effect data, and user offline service feedback;

[0038] A time-series analysis of the offline data of the stores is performed to obtain a set of offline indicators for the stores, which includes real-time service carrying capacity indicators, offline product conversion rate indicators, and user experience satisfaction indicators.

[0039] Obtain the user's offline data generated by the user in at least two different recommended stores for the same product. The user's offline data includes the user's browsing trajectory, trial records, offline transaction data, store service evaluations and membership points usage records. Compare the user's offline data generated for the same product to obtain the difference store offline index.

[0040] Based on the first store profile, the store offline indicator set, and the difference store offline indicator, a second store profile is obtained. The second store profile includes the store's response speed to offline activities, resource allocation efficiency matrix, and user experience optimization path.

[0041] Preferably, the filtering based on the second user profile within the second store profile includes the following steps:

[0042] By analyzing the second user profile and each of the resource allocation efficiency matrices, a store fit score is obtained. The fit score is used to quantify the store's ability to meet the user's offline needs.

[0043] The stores are then ranked using a secondary weighted index based on the aforementioned user experience satisfaction metrics.

[0044] Stores that do not meet the requirements are filtered out by dynamic constraints, which include the store's current customer traffic threshold, the offline inventory replenishment cycle of goods, and the user's specified time window for visiting the store.

[0045] Preferably, obtaining the sorted list of recommended stores in response to the second user profile includes the following steps:

[0046] A multi-objective optimization model is constructed, wherein the objective function is to maximize the store fit score, minimize the user waiting time, and maximize the offline conversion rate of the product.

[0047] Based on the results of the multi-objective optimization model, the priority ranking of the recommended stores is obtained;

[0048] For each store under the priority ranking, an offline scenario-based recommendation scheme is generated, which includes store-exclusive discount combinations, personalized service recommendations, and in-store route planning.

[0049] The sorted list is obtained based on the priority ranking of the recommended stores and the offline scenario-based recommendation scheme of the recommended stores;

[0050] The sorted list is pushed to the user's terminal and synchronized to the corresponding recommended stores;

[0051] When a new store that does not belong to the recommended stores appears in the results of the multi-objective optimization model, the new store and its corresponding offline scenario-based recommendation scheme are pushed to the user terminal and synchronized to the corresponding new store.

[0052] Beneficial Effects: The AI-based multi-store O2O smart mall system of this application integrates online data to generate a first user profile through a user profiling module, and updates it to a second user profile by combining offline data, comprehensively capturing changes in user needs; the store profiling module integrates online and offline data to construct store profiles, accurately reflecting operational quality and supply capacity; the recommendation module dynamically filters and recommends stores based on profiles and updates them in real time, combining multi-objective optimization to generate scenario-based solutions; it achieves precise matching between users and stores, improves the accuracy and timeliness of recommendations, optimizes the offline user experience, increases purchase conversion rates, and at the same time helps stores efficiently allocate resources, enhancing the operational efficiency and commercial value of the multi-store O2O model. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 The structural block diagram of the AI-based multi-store O2O smart mall system provided in the embodiments of this application. Detailed Implementation

[0055] The technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0056] In this document, the term "comprising" is intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0057] In the operation scenario of multi-store O2O (online-to-offline) smart malls, accurately matching user needs with store service capabilities is the core issue for improving user experience and operational efficiency. Current technologies often limit user profiling to online data, failing to reflect actual offline user needs; store information is presented statically, not fully reflecting operational quality and supply capacity; and recommendation results lack a real-time adjustment mechanism based on dynamic offline behavior, resulting in insufficient recommendation accuracy and low user-store matching.

[0058] To address the aforementioned issues, this invention provides an AI-based multi-store O2O smart mall system that achieves accurate filtering and real-time optimization of recommended stores and products by constructing dynamically updated user profiles and store profiles.

[0059] Example 1

[0060] like Figure 1 As shown, an AI-based multi-store O2O smart mall system includes:

[0061] The user profile module is used to obtain the first user profile through online user data. The first user profile is used to define the user's online intention for products.

[0062] The store profile module is used to obtain the first store profile through online store data. The first store profile is used to define the store's operating quality and supply capacity.

[0063] The recommendation module is used to filter from the first user profile and the first store profile to obtain at least one recommended store that responds to the first user profile; wherein, the recommended stores are updated in real time based on the offline activities of users and stores.

[0064] It should be noted that in this embodiment, existing AI (Artificial Intelligence) profiling technology is used to construct the first user profile and the first store profile. Furthermore, the user profile module can be located on the user terminal, which hosts a corresponding e-commerce platform. The store profile module is located on the operating system of the store. Both modules are communicatively connected to the recommendation module, which can be located on the server of the AI-based multi-store O2O smart mall system.

[0065] By leveraging the above, precise online matching of users and stores can be achieved through the first user profile and the first store profile, thereby improving the accuracy of store and product recommendations.

[0066] In the specific implementation of building online user profiles, to comprehensively capture users' potential needs, this embodiment adopts existing multi-dimensional data fusion and machine learning analysis methods. By collecting users' complete behavioral trajectories on existing e-commerce platforms, including browsing, searching, favorites, adding to cart, transactions, and reviews, and combining existing feature engineering and clustering algorithms, discrete user behaviors are transformed into a quantifiable preference vector space.

[0067] Specifically, the first user profile is obtained through online user data, including the following steps:

[0068] Acquire user online data, which includes users' online browsing history, search keywords, favorited products, add-to-cart behavior, historical order data, and review data;

[0069] Feature extraction and cluster analysis are performed on user online data to obtain corresponding preference vectors, which include product category preference, price sensitivity, brand preference, and purchase frequency characteristics.

[0070] The first user profile is obtained based on the preference vector. The first user profile includes the user's interest rating for different categories of goods and the prediction of potential demand.

[0071] Based on the above, this embodiment obtains a first user profile based on online user data, and achieves accurate definition of user interest ratings and potential demand prediction for different categories of goods.

[0072] To achieve a digital representation of store service capabilities, this embodiment adopts a technical solution of constructing a store profile system, which quantifies store operational quality and supply capacity by integrating multi-source data. After acquiring online store data, heterogeneous data is standardized using existing data processing technologies to establish a unified evaluation index system, thereby transforming store characteristics such as merchandise inventory, fulfillment efficiency, and service quality into a computable matrix model.

[0073] Specifically, by using online store data, we obtain the first store profile, which includes the following steps:

[0074] Acquire online data from stores, including store inventory data, historical order fulfillment rate, average delivery time, user review data, promotional activity records, and regional coverage.

[0075] The online data of the stores is standardized to obtain a set of online indicators for the stores, which includes indicators of the store's product richness, service quality and fulfillment capabilities.

[0076] The first store profile is obtained based on the online indicator set of the store. The first store profile includes the store's operational quality scoring system and the product supply capacity feature matrix.

[0077] Based on the above, this embodiment obtains a first store profile based on online store data, which enables an accurate definition of the store's operational quality and supply capacity. The resulting store operational quality scoring system and product supply capacity feature matrix provide a data foundation for product recommendations at the store level.

[0078] In the intelligent matching process between users and stores, this embodiment employs a multi-stage screening mechanism, achieving accurate recommendations through similarity calculation, weighted ranking, and dynamic constraint filtering. This process considers both the matching degree between user preferences and store supply, prioritizes store operating quality, and ensures the feasibility of the recommendation results through constraints.

[0079] Specifically, filtering is performed based on the first user profile and the first store profile, including the following steps:

[0080] The similarity score is calculated by comparing the first user profile with the product supply capacity feature matrix. The similarity score is used to define the store's corresponding capacity for the first user profile.

[0081] Stores are ranked by weight based on an operational quality scoring system;

[0082] Stores that do not meet the requirements are filtered out by constraints, including user-specified delivery range, expected delivery time, and specific product inventory thresholds.

[0083] In this embodiment, the weights in the weighted ranking of stores based on the operational quality scoring system can be the weights corresponding to product matching degree, delivery timeliness, service quality and inventory adequacy rate, and the specific weight values ​​can be set based on actual needs.

[0084] Based on the above, this embodiment uses the first user profile to filter the first store profile, providing a data foundation for generating recommended stores.

[0085] Furthermore, to enable personalized recommendations, after selecting the corresponding stores, customized product combinations and service plans are generated for them. This process combines user interest scores with real-time store inventory dynamics, ensuring the availability of recommended products offline while achieving online-offline service synergy through simultaneous push notifications across multiple terminals.

[0086] Specifically, obtaining at least one recommended store that responds to the first user profile includes the following steps:

[0087] Output stores with similarity scores greater than the preset similarity score threshold and business quality scores in the top N, obtain a store list, and determine the stores in the store list as recommended stores;

[0088] For each store in the store list, a personalized product recommendation combination is generated. The product combination is generated based on the store's inventory and the user's interest rating for each product category. The personalized product recommendation combination is pushed to the user's terminal, and the personalized product recommendation combination and its corresponding user information are pushed to the recommending store.

[0089] It should be noted that in this embodiment, the similarity score threshold is obtained by fitting historical similarity scores and their corresponding store selections; while the top N stores are set based on actual needs, but at least two stores should be recommended to provide users with more options.

[0090] Based on the above, this embodiment provides users with at least two precise selection options by obtaining at least one recommended store that responds to the first user profile, and provides stores with precise user information.

[0091] In Example 1, the user profiling module integrates online multi-dimensional behavioral data to generate accurate user profiles, and combines this with the store profiling module to quantitatively analyze the store's operational quality and supply capacity, achieving precise matching between user needs and store services. The recommendation module uses similarity calculation, weighted sorting, and constraint filtering to select suitable recommended stores and generate personalized product combinations that are synchronized to both users and stores, improving the accuracy and timeliness of recommendations. This effectively solves the problems of incomplete user profiles, neglect of store differences, and disconnected recommendations in the traditional O2O model, optimizes user experience, increases purchase conversion rates, and enhances the efficiency of multi-store collaborative operations.

[0092] Example 2

[0093] After a user visits a recommended store for an offline activity, their needs and preferences, as well as the store's service capabilities, may dynamically change based on the actual interaction. To achieve closed-loop optimization of recommendations, this embodiment further integrates online and offline data to dynamically update user and store profiles, and regenerates the recommendation ranking based on the updated profiles.

[0094] Specifically, when a user visits different recommended stores, the user profile module obtains a second user profile through the user's offline data. The second user profile is used to define the changes in the user's intention to purchase products in the first user profile after the offline activity.

[0095] The store profile module is used to obtain a second store profile through offline store data. The second store profile is used to define how recommended stores respond to users' offline activities.

[0096] The recommendation module is used to filter through the second user profile and the second store profile to obtain a sorted list of recommended stores in response to the second user profile.

[0097] It should be noted that in this embodiment, the construction technology of the second user profile and the second store profile is the same as that of the first user profile and the first store profile in Embodiment 1. The difference lies in the data used during construction.

[0098] Based on the above, this embodiment uses a user profile module, a store profile module, and a recommendation module to update the first user profile and the first store profile based on offline user data and offline store data, thereby obtaining a second user profile and a second store profile, providing a technical foundation for further improving the processing quality of multi-store O2O.

[0099] To accurately capture the impact of users' offline behavior on demand, differentiated analysis of offline user data across multiple stores is necessary. By extracting user behavior patterns and interaction depths across different stores, and comparing these with the initial online user profile, the patterns of interest changes can be quantified, thereby enabling offline adjustments to the user profile.

[0100] Specifically, the second user profile is obtained through offline user data, including the following steps:

[0101] Acquire offline user data generated at recommended stores. Offline user data includes browsing history, trial records, offline transaction data, store service evaluations, and membership points usage records.

[0102] Spatiotemporal features are extracted from offline user data to construct offline behavior sequences, including the duration of user stay in each store, depth of product interaction, and decision-making path.

[0103] By comparing and analyzing offline user data with the first user profile, we can obtain the rate of change in user interest in different product categories. The rate of change in interest includes positive enhancement factors and negative decay factors.

[0104] The potential demand prediction is updated based on the interest change rate, and a second user profile is generated. The second user profile includes the user's preference correction coefficient matrix in offline scenarios.

[0105] It should be noted that this embodiment acquires user offline data based on existing data collection methods, such as video data collection by intelligent robots set up in offline stores or data collection systems deployed in the operating system of offline stores; it constructs user offline behavior sequences based on existing spatiotemporal feature extraction technology, and obtains the interest change rate and updates potential demand predictions based on the interest change rate based on existing data processing technology.

[0106] In practice, a store's responsiveness to offline activities is dynamic and requires analysis of real-time operational data and cross-store comparisons to update the store profile. This embodiment collects offline data such as real-time customer traffic and promotional effectiveness to construct an evaluation system that includes indicators such as service capacity and conversion rate. It also combines this data with differences in user behavior across multiple stores to generate a second store profile that reflects the offline service characteristics of each store.

[0107] Specifically, by using offline store data, a second store profile is created, including the following steps:

[0108] Obtain offline data from each store, including real-time customer flow data, merchandise display adjustment records, offline promotional activity effectiveness data, and user feedback on offline services.

[0109] Time-series analysis of offline store data yields a set of offline store metrics, including real-time service capacity metrics, offline product conversion rate metrics, and user experience satisfaction metrics.

[0110] The system acquires offline user data generated by users at at least two different recommended stores for the same product. This offline user data includes browsing history, trial records, offline transaction data, store service reviews, and membership points usage records. The system compares the offline user data generated for the same product to obtain the difference store offline indicator. In this embodiment, the difference store offline indicator is obtained based on existing data processing technologies. For example, data fusion is used to fuse the differences in multi-dimensional offline user data to obtain the difference store offline indicator.

[0111] Based on the first store profile, the store's offline indicator set, and the difference store's offline indicator, a second store profile is obtained. The second store profile includes the store's response speed to offline activities, resource allocation efficiency matrix, and user experience optimization path.

[0112] Based on the above, this embodiment obtains a second store profile based on the first store profile, the store offline indicator set, and the differential store offline indicator, so that the second store profile fully describes the store's operational quality and supply capacity. Among them, the differential store offline indicator directly defines the store's operational capacity for the same product, providing data reference for multi-store managers to manage multiple stores, and also providing data reference for store operators to improve the quality of store operations.

[0113] When filtering based on the updated second user profile and second store profile, it is necessary to consider the dynamic matching between user offline preferences and store real-time service capabilities. This embodiment ensures that the recommendation results are highly consistent with the offline scenario through quantitative fit, dynamic weighted ranking, and dynamic constraint filtering.

[0114] Specifically, filtering is performed using the second user profile and the second store profile, including the following steps:

[0115] By analyzing the second user profile and the resource allocation efficiency matrix, a store fit score is obtained. The fit score is used to quantify the store's ability to meet users' offline needs.

[0116] The stores were then ranked using a secondary weighted ranking based on user experience satisfaction metrics.

[0117] Stores that do not meet the requirements are filtered out by dynamic constraints, which include the store's current customer traffic threshold, the offline inventory replenishment cycle of goods, and the user's specified time window for visiting the store.

[0118] In this embodiment, the weights for secondary weighting of stores based on user experience satisfaction indicators may include offline service response speed, offline product display effect, and personalized service matching degree, and the specific weight values ​​can be set based on actual needs.

[0119] Based on the above, filtering through the second user profile and the second store profile provides a technical foundation for further optimizing product recommendations.

[0120] To generate a recommended ranking list that aligns with users' real-time needs and the real-time service capabilities of stores, a multi-objective optimization model is needed to balance fit, efficiency, and conversion rate, while also incorporating scenario-based solutions to enhance the offline experience. Additionally, it should support the inclusion of new stores to expand the recommendation scope.

[0121] Specifically, obtaining a sorted list of recommended stores in response to the second user profile includes the following steps:

[0122] A multi-objective optimization model is constructed, with the objective functions being to maximize the store fit score, minimize user waiting time, and maximize the offline conversion rate of goods.

[0123] Based on the results of the multi-objective optimization model, the priority ranking of recommended stores is obtained;

[0124] For each store ranked by priority, generate an offline scenario-based recommendation plan, which includes store-exclusive discount combinations, personalized service recommendations, and in-store traffic flow planning.

[0125] A sorted list is obtained based on the priority ranking of recommended stores and the offline scenario-based recommendation scheme for each recommended store;

[0126] The sorted list is pushed to the user's device and synchronized to the corresponding recommended stores;

[0127] When a new store that is not among the recommended stores appears in the results of the multi-objective optimization model, the new store and its corresponding offline scenario-based recommendation solution are pushed to the user's terminal and synchronized to the corresponding new store.

[0128] In this embodiment, existing multi-objective optimization model technology is used to maximize store fit score, minimize user waiting time, and maximize offline product conversion rate.

[0129] In Example 2, a second user profile is generated by integrating offline user data from multiple stores through a user profiling module, accurately capturing changes in user needs. A second store profile is constructed by combining real-time customer flow, promotional effectiveness, and cross-store comparisons to reflect service responsiveness. The recommendation module generates a recommendation ranking list based on the updated profiles, through adaptation analysis, dynamic sorting, and multi-objective optimization, simultaneously implementing scenario-based solutions. This achieves dynamic and accurate matching between users and stores, improves offline experience and conversion rates, supports new store expansion, solves the problems of lagging recommendations and insufficient utilization of offline data in traditional O2O, and enhances the flexibility and commercial value of multi-store collaborative operations.

[0130] In summary, the AI-based multi-store O2O smart mall system in this embodiment integrates online data to generate a first user profile through a user profiling module, and updates it to a second user profile by combining offline data, comprehensively capturing changes in user needs. The store profiling module integrates online and offline data to construct store profiles, accurately reflecting operational quality and supply capacity. The recommendation module dynamically filters and recommends stores based on the profiles and updates them in real time, combining multi-objective optimization to generate scenario-based solutions. This achieves precise matching between users and stores, improves the accuracy and timeliness of recommendations, optimizes the offline user experience, increases purchase conversion rates, and helps stores efficiently allocate resources, enhancing the operational efficiency and commercial value of the multi-store O2O model.

[0131] In the embodiments provided in this application, it should be understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any suitable combination thereof. For hardware implementation, the processor may be implemented in one or more of the following: application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, other electronic units designed to implement the functions described herein, or combinations thereof. For software implementation, some or all of the processes of the embodiments may be performed by a computer program instructing the associated hardware. During implementation, the program may be stored in a computer-readable storage medium or transmitted as one or more instructions or code on a computer-readable storage medium. Computer-readable storage media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of a computer program from one place to another. Storage media may be any available medium accessible to a computer. Computer-readable storage media may include, but are not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code having the form of instructions or data structures and accessible to a computer.

[0132] Finally, it should be noted that the above description is only a preferred embodiment of this application and is not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A multi-store O2O smart mall system based on AI, characterized in that, The system includes: The user profile module is used to obtain the first user profile through online user data. The first user profile is used to define the user's online intention for products. The store profile module is used to obtain the first store profile through online store data. The first store profile is used to define the store's operating quality and supply capacity. The recommendation module is used to filter the first store profile based on the first user profile to obtain at least one recommended store that responds to the first user profile; wherein the recommended stores are updated in real time based on the offline activities of users and stores.

2. The AI-based multi-store O2O smart mall system according to claim 1, characterized in that, The process of obtaining the first user profile through online user data includes the following steps: The user's online data is acquired, including the user's online browsing history, search keywords, favorited products, add-to-cart behavior, historical order data, and review data. Feature extraction and cluster analysis are performed on the user's online data to obtain the corresponding preference vector, which includes product category preference, price sensitivity, brand preference, and purchase frequency characteristics. The first user profile is obtained based on the preference vector. The first user profile includes the user's interest rating for different categories of goods and prediction of potential demand.

3. The AI-based multi-store O2O smart mall system according to claim 1, characterized in that, The process of obtaining the first store profile through online store data includes the following steps: The online data of the stores is obtained, including the stores' product inventory data, historical order fulfillment rate, average delivery time, user review data, promotional activity records, and regional coverage. The online data of the stores is standardized to obtain a set of online indicators for the stores, which includes indicators of the store's product richness, service quality, and fulfillment capability. The first store profile is obtained based on the online indicator set of the store. The first store profile includes the store's operational quality scoring system and commodity supply capacity feature matrix.

4. The AI-based multi-store O2O smart mall system according to claim 3, characterized in that, The step of filtering within the first store profile based on the first user profile includes the following steps: The similarity score is calculated by comparing the first user profile with the product supply capacity feature matrix. The similarity score is used to define the store's corresponding capacity for the first user profile. The stores are ranked using a weighted average based on the aforementioned operational quality scoring system; Stores that do not meet the requirements are filtered out by constraints, including user-specified delivery range, expected delivery time, and specific product inventory thresholds.

5. The AI-based multi-store O2O smart mall system according to claim 4, characterized in that, Obtaining at least one recommended store in response to the first user profile includes the following steps: Output the stores whose similarity scores are greater than a preset similarity score threshold and whose business quality scores are in the top N, and obtain a store list. Then, determine the stores in the store list as the recommended stores. For each store in the store list, a personalized recommended product combination is generated. The product combination is generated based on the store's inventory and the user's interest rating for each product category. The personalized recommended product combination is pushed to the user's terminal, and the personalized recommended product combination and its corresponding user information are pushed to the recommended store.

6. The AI-based multi-store O2O smart mall system according to claim 2, characterized in that, When a user visits different recommended stores, the user profiling module obtains a second user profile through the user's offline data. The second user profile is used to define the changes in the user's intention to purchase products in the first user profile after the offline activity. The store profile module is used to obtain a second store profile through offline store data. The second store profile is used to define the response of the recommended store to the user's offline activities. The recommendation module is used to filter the second store profile based on the second user profile to obtain a sorted list of recommended stores in response to the second user profile.

7. The AI-based multi-store O2O smart mall system according to claim 6, characterized in that, The process of obtaining the second user profile through offline user data includes the following steps: The system acquires the user's offline data generated at the recommended stores, including browsing history, trial records, offline transaction data, store service evaluations, and membership points usage records. Spatiotemporal features are extracted from the user's offline data to construct the user's offline behavior sequence, which includes the user's dwell time in each store, product interaction depth, and decision path. By comparing and analyzing the user's offline data with the first user profile, the change rate of user interest in different product categories is obtained. The change rate of interest includes a positive increase factor and a negative decrease factor. The potential demand prediction is updated based on the interest change rate, and a second user profile is generated. The second user profile includes a preference correction coefficient matrix for users in offline scenarios.

8. The AI-based multi-store O2O smart mall system according to claim 6, characterized in that, The process of obtaining the second store profile through offline store data includes the following steps: Obtain the offline data of each store, including real-time customer flow data, merchandise display adjustment records, offline promotional activity effect data, and user offline service feedback; A time-series analysis of the offline data of the stores is performed to obtain a set of offline indicators for the stores, which includes real-time service carrying capacity indicators, offline product conversion rate indicators, and user experience satisfaction indicators. Obtain the user's offline data generated by the user in at least two different recommended stores for the same product. The user's offline data includes the user's browsing trajectory, trial records, offline transaction data, store service evaluations and membership points usage records. Compare the user's offline data generated for the same product to obtain the difference store offline index. Based on the first store profile, the store offline indicator set, and the difference store offline indicator, a second store profile is obtained. The second store profile includes the store's response speed to offline activities, resource allocation efficiency matrix, and user experience optimization path.

9. The AI-based multi-store O2O smart mall system according to claim 8, characterized in that, The filtering process based on the second user profile within the second store profile includes the following steps: By analyzing the second user profile and each of the resource allocation efficiency matrices, a store fit score is obtained. The fit score is used to quantify the store's ability to meet the user's offline needs. The stores are then ranked using a secondary weighted index based on the aforementioned user experience satisfaction metrics. Stores that do not meet the requirements are filtered out by dynamic constraints, which include the store's current customer traffic threshold, the offline inventory replenishment cycle of goods, and the user's specified time window for visiting the store.

10. The AI-based multi-store O2O smart mall system according to claim 9, characterized in that, Obtaining the sorted list of recommended stores in response to the second user profile includes the following steps: A multi-objective optimization model is constructed, wherein the objective function is to maximize the store fit score, minimize the user waiting time, and maximize the offline conversion rate of the product. Based on the results of the multi-objective optimization model, the priority ranking of the recommended stores is obtained; For each store under the priority ranking, an offline scenario-based recommendation scheme is generated, which includes store-exclusive discount combinations, personalized service recommendations, and in-store route planning. The sorted list is obtained based on the priority ranking of the recommended stores and the offline scenario-based recommendation scheme of the recommended stores; The sorted list is pushed to the user's terminal and synchronized to the corresponding recommended stores; When a new store that does not belong to the recommended stores appears in the running results of the multi-objective optimization model, the new store and its corresponding offline scenario-based recommendation scheme are pushed to the user terminal and synchronized to the corresponding new store.

Images