Commodity recommendation method and device for intelligent shopping cart and storage medium

Abstract

The invention discloses a commodity recommendation method and device for an intelligent shopping cart and a storage medium, and belongs to the technical field of intelligent shopping carts, and the method comprises the steps: obtaining the real-time position of a user in a supermarket and supermarket and the hand motion data of the user; based on the real-time position and the hand action data, determining an effective stopping duration; based on the effective stopping duration, constructing or updating a score set of the commodities corresponding to the real-time positions by the users, and calculating user similarity through a collaborative filtering algorithm based on the users to obtain a complete score set; generating first m commodity recommendation lists with the highest predicted scores in the current area according to the complete score set and the condition that the user staying duration reaches a preset triggering condition, and performing real-time recommendation display through an interactive interface; according to the method, Bluetooth positioning and an improved collaborative filtering algorithm are combined, a relatively delicate intelligent recommendation function is achieved, and compared with a relatively single mode of recommending homogenized commodities and introducing partner commodity concepts, a guiding effect on a user purchase behavior is completed.

Description

Technical Field

[0001] The present invention relates to a product recommendation method, device and storage medium for an intelligent shopping cart, belonging to the technical field of intelligent shopping carts. Background Art

[0002] With the development of the times and technology, in the offline retail industry, intelligent shopping carts have become a new trend. Traditional shopping carts are merely a vehicle for goods. Intelligent shopping carts have realized an intelligent and automated shopping method through gravity or in combination with image algorithms, and support automatic checkout and real-time viewing of the total amount of goods already added to the shopping cart. These functions have greatly solved the problem of queuing in large and popular shopping malls during holidays. However, under the domestic cultural system, most people go to shopping malls without a clear shopping goal. Then, there is no such intelligent recommendation algorithm on existing intelligent shopping carts to help customers accurately recommend products during the process of wandering around.

[0003] Existing mature indoor Bluetooth positioning algorithms mainly calculate the time of arrival of the received signal between the base station and the mobile station, and then convert it into distance for positioning. This method requires at least three Bluetooth beacons to calculate the position of the target.

[0004] Existing collaborative filtering algorithms are divided into two types. One is the user-based collaborative filtering algorithm, which calculates the similarity between users, and then predicts the degree of interest of the target user in unrated items based on the behaviors and ratings of similar users. The other is the item-based collaborative filtering algorithm, which calculates the similarity between different items, and then predicts the degree of interest of the target user in unrated items based on the ratings of the target user for similar items. There are some flaws in applying this algorithm to intelligent shopping carts. The first is that users will not rate products in the shopping cart. The second is that this algorithm essentially always recommends similar products to customers. In the field of short videos or online novels, continuously recommending homogenized content, users can still pay more attention due to personal preferences. However, in shopping, users rarely buy similar products or even products of the same category but only different brands. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art, and provide a product recommendation method, device and storage medium for an intelligent shopping cart. By combining Bluetooth positioning with an improved collaborative filtering algorithm, a more refined intelligent recommendation function is achieved. Compared with simply recommending homogenized products, the concept of partner products is introduced to complete the guiding role for users' purchase behaviors.

[0006] To achieve the above purpose, the present invention is implemented by adopting the following technical solutions: In a first aspect, the present invention provides a product recommendation method for an intelligent shopping cart, including: Obtaining the real-time position of the user in the shopping mall through Bluetooth beacon positioning, and obtaining the user's hand movement data through a vision detection module; Based on the real-time position and hand movement data, detecting whether the user's stay time at the current position exceeds a preset threshold, and combining a target detection algorithm to determine whether the user takes out a product; if both the stay time exceeds the threshold and the taking-out action is detected, it is recorded as the effective stay duration; Based on the effective stay duration, constructing or updating the scoring set of the user for the products corresponding to the real-time position. Taking the scoring set as the input, calculating the user similarity through a user-based collaborative filtering algorithm to obtain the scoring set of unrated products, so as to obtain the complete scoring set of the user for the products in the area; According to the complete scoring set and the user's stay duration reaching a preset trigger condition, generating a recommended list of the top m products with the highest predicted scores in the current area, and recommending and displaying them in real time through an interactive interface; When the user scans the code to add a product to the cart, based on the complete scoring set and the pre-acquired historical purchase data, calculating the partner threshold set and the associated product scores of the scanned product through an item-based collaborative filtering algorithm, generating an associated product recommended list, and recommending and displaying it in real time through an interactive interface.

[0007] Further, the calculation formula of the user similarity is as follows: S(u,v)= ; Wherein, S(u,v) represents the user similarity, represents the product, represents the user 's scoring set, represents the user 's evaluation set, represents the user's score for for 's score, represents the user for 's score, and is mapped and generated by the effective stay duration, represents the user 's average stay duration, ` represents the user 's average stay duration.

[0008] Further, the score of the unrated product is calculated as follows: ; Wherein, Represents the predicted user For the commodity Rating of Represents the neighbor set Represents the user's rating of For the commodity Rating of

[0009] Furthermore, the calculating of the partner threshold set and the associated commodity rating of the scanned commodity by the item-based collaborative filtering algorithm includes: Calculating the partner threshold of the scanned commodity , the formula is as follows: ; Wherein, Represents the number of users who purchase Item Represents the number of users who purchase Item Represents the number of items that are liked by both And Item And Generated based on the user's historical purchase data; Calculating the associated commodity rating through the partner threshold of the commodity, the formula is as follows: ; Wherein, Represents the user For Predicted rating of the commodity Represents the set of commodities that the user has historically purchased and settled Represents the scanned Commodity with the highest partner threshold Items

[0010] Furthermore, the method further includes: determining whether the user is in the store for the first time. If the user is a first-time user, then recommend the commodity with the highest rating in the current area.

[0011] Furthermore, the method further includes: when recommending and displaying in real time through the interactive interface, supporting the feedback of "not interested" through the interactive button to optimize the recommendation result. When the user clicks "not interested", exclude the corresponding commodity in real time and update the recommendation list.

[0012] Furthermore, the method further includes: when recommending associated commodities, preferentially display the partner commodities with targeted discounts for the current commodity, and guide the user to quickly locate the recommended commodities through the shopping cart navigation function.

[0013] In a second aspect, the present invention provides a commodity recommendation device for an intelligent shopping cart, including: A data acquisition module, configured to obtain the real-time location of a user in a shopping mall through Bluetooth beacon positioning, and obtain the user's hand movement data through a vision detection module; A first data processing module, configured to detect whether the residence time of the user at the current location exceeds a preset threshold based on the real-time location and the hand movement data, and combine a target detection algorithm to determine whether the user takes out a commodity; if both the residence time exceeds the threshold and the taking-out action is detected, it is recorded as the effective residence duration; A second data processing module, configured to construct or update a rating set of the commodity corresponding to the real-time location by the user based on the effective residence duration, take the rating set as an input, calculate the user similarity through a collaborative filtering algorithm based on the user, obtain a rating set of unrated commodities, so as to obtain a complete rating set of the commodities in the area by the user; A first recommendation module, configured to generate a recommended list of the top m commodities with the highest predicted ratings in the current area according to the complete rating set and the user residence duration reaching a preset trigger condition, and recommend and display them in real time through an interaction interface; A second recommendation module, configured to, when the user scans the code to add a commodity to the shopping cart, calculate a partner threshold set and an associated commodity rating of the scanned commodity through a collaborative filtering algorithm of items based on the complete rating set and the pre-obtained historical purchase data, generate a recommended list of associated commodities, and recommend and display them in real time through an interaction interface.

[0014] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the method described in any one of the foregoing are implemented.

[0015] In a fourth aspect, the present invention provides a computer device, including: A memory, configured to store a computer program / instructions; A processor, configured to execute the computer program / instructions to implement the steps of the method described in any one of the foregoing.

[0016] In a fifth aspect, the present invention provides a computer program product, including a computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the method described in any one of the foregoing are implemented.

[0017] Compared with the prior art, the beneficial effects achieved by the present invention: The present invention provides a product recommendation method, device and storage medium for an intelligent shopping cart. By combining Bluetooth positioning with an improved collaborative filtering algorithm, a relatively sophisticated intelligent recommendation function is achieved. Compared with simply recommending homogeneous products, the concept of partner products is introduced to guide users' purchasing behavior. For users, the first-step recommendation function can conveniently lock in the initial purchase target. For the supermarket, the second-step partner recommendation function can increase the time users stay in the supermarket and stimulate their purchasing desire as much as possible. Brief Description of the Drawings

[0018] Figure 1 It is a component structure diagram of a product recommendation method for an intelligent shopping cart provided by an embodiment of the present invention; Figure 2 It is a flowchart of a product recommendation method for an intelligent shopping cart provided by an embodiment of the present invention. Detailed Embodiments

[0019] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be used to limit the protection scope of the present invention.

[0020] Embodiment 1. This embodiment introduces a product recommendation method for an intelligent shopping cart, including: Obtain the real-time position of the user in the supermarket through Bluetooth beacon positioning, and obtain the user's hand movement data through a visual detection module; Based on the real-time position and hand movement data, detect whether the user's stay time at the current position exceeds a preset threshold, and combine a target detection algorithm to determine whether the user takes out a product; if both the stay time exceeds the threshold and the taking-out action is detected, record it as the effective stay duration; Based on the effective stay duration, construct or update the user's score set for the products corresponding to the real-time position. Taking the score set as the input, calculate the user similarity through a user-based collaborative filtering algorithm to obtain the score set of unrated products, so as to obtain the complete score set of the user for the products in the area; According to the complete score set and the user's stay duration reaching a preset trigger condition, generate a recommended list of the top m products with the highest predicted scores in the current area, and display the real-time recommendation through an interaction interface; When the user scans the code to add a product to the cart, based on the complete score set and the previously obtained historical purchase data, calculate the partner threshold set and the associated product scores of the scanned product through an item-based collaborative filtering algorithm, generate an associated product recommendation list, and display the real-time recommendation through an interaction interface.

[0021] The product recommendation method for an intelligent shopping cart provided in this embodiment specifically involves the following steps in its application process: First, the scope of the recommendation algorithm is a single shopping mall. This is because of the differences in users' shopping habits due to different regions. Then, in the user-based collaborative filtering algorithm mentioned above, the rating of a user for a product is converted into the time the user stays in front of a product, which is calculated through indoor Bluetooth positioning. In a shopping mall, the placement positions of single products are relatively concentrated and fixed. By regularly receiving and updating the position map of the products placed in the shopping mall, it is possible to accurately know which products are in a certain area. Then, when the user stays there, their indoor Bluetooth positioning remains unchanged. To prevent device errors, it can be set that only when the staying time is greater than 3s is it recorded as valid data. Secondly, in order to accurately know that the staying time of the user in this product area is valid staying time (valid staying time means that the user stays here to actually browse products rather than chatting with acquaintances or standing and resting while playing with the mobile phone by chance), according to on-site investigation and big data analysis, most users will take the products they are interested in out of the shelf and hold them in their hands to observe. Then, camera probes are set up at the top of the shelf to capture the user's hand movement behavior. The short video captured (which can be 3s or 5s) is transmitted to the central server. The server uses the YOLO algorithm (object detection algorithm) to determine whether the user takes out the product event. If it is a product-taking-out event, a signal is sent to the RFID electronic tag bound to the vehicle (each intelligent shopping cart comes with an RFID for querying the settlement completion information when completing the settlement and opening the exit gate according to this information). The calculation of the effective staying duration starts from this moment and ends when the user moves. When entering the next stay, the above process is repeated. In this way, through algorithm calculation, the approximate ratings of all products in the shopping mall by the user can be obtained, and the ratings are reflected by the length of the staying time.

[0022] Secondly, the concept of partner recommendation is introduced. Partner recommendation means related products. For example, when buying instant noodles, the partners of this product are usually pickled eggs and ham sausages. This move aims to solve the problem that the algorithm still recommends instant noodles even when instant noodles have been added to the shopping cart, and hopes to recommend products that can be used in combination or as partners to the user. Here, the item-based collaborative filtering algorithm is used to calculate the similarity between different items in a single shopping mall (such as between product x and product y). Because in this algorithm, essentially when user A buys product x and product y, user b who buys product x is recommended to buy product y. Then, the obtained product similarity can be regarded as the partner threshold of product y for product x. In this way, through the calculation of a large sample, the collection of partner thresholds of other products in the shopping mall for product x can be obtained. To save performance and storage space, it can be flexibly adjusted based on the UI design (for example, if the system interface design can only display 4 recommendation positions, then only the top four highest partner thresholds of a single product need to be recorded).

[0023] Now, based on the above two steps, we can obtain the product rating set of a single user for the supermarket and the partner threshold of a single product for other products in the supermarket. Additionally, the sojourn time t0 is defined here. When the duration that a user stays in a certain area is greater than t0, the user has a relatively high purchase propensity for the products in this area. At this time, recommend to the user the m products with the highest ratings predicted by the algorithm in this area. There is a situation where user c comes to this supermarket for the first time. For c, his rating set is empty or the total number of his rating set predicted by the algorithm is less than m. Then the first recommendation is the m products with the best ratings of all users in this area. For c, his recommendations will become more and more accurate after visiting this supermarket multiple times.

[0024] When the user scans the code to add a product to the cart, the system will recommend the n products with the highest partner thresholds calculated based on the scanned product. Here, products with targeted discounts will be recommended first (if a single product has a discount and its partner threshold is among the top n, then it will be preferentially displayed at the top or on the first page). Combining with the navigation function of the shopping cart, users can quickly find the partner products they want to add to the cart. At the same time, considering the situation where users are not interested, a page interaction button "Not interested" is designed. Then this product will not be displayed in this recommendation.

[0025] Figure 1 To form a structure diagram, where Bluetooth indoor positioning provides calculation parameters for the user-based collaborative filtering algorithm (User-basedCF). The implementation principle is to convert the user's ratings of items into the sojourn time in front of the products, and specify the time to start the recommendation function based on the time the user lingers. Moreover, based on the item-based collaborative filtering algorithm (Items-based CF), the partner thresholds between different products are calculated to meet the diversified recommendation functions and the needs of users for combined use. The existing intelligent shopping system provides the function and platform for adding products to the cart. Relying on this function, the timing and display method of recommending partner products are specified. Among them, User-based CF uses the Pearson correlation coefficient for calculating the similarity between users. The calculation formula is: S(u,v)= (1); Among them, S(u,v) represents the similarity, represents the product, represents the user 's rating set, represents the user 's evaluation set, represents the user's rating of for , represents the user 's rating of for and is generated by mapping the effective stay duration, representing the average stay duration of the user `, indicating the average stay duration of the user After the above steps, the similarity S(u, v) between users u and v is obtained, and then the predicted score of user u for the unrated items is calculated. The calculation formula used here is: (2); indicating the predicted score of user for item `, denotes the neighbor set, which represents a set to which two users calculating similarity belong. The existing product functions support collecting store self-portraits and generating user groups. After user classification, can be positioned into different user groups (shopping habits of different groups are different), indicating the score of user for for item `. Repeating the above process to obtain the predicted scores of user for the unrated items in the set G (G represents a certain item set), a set indicating the predicted score set of user can be obtained. Combining with the existing score set, the score set of the user for the items in this area can be obtained, and the top m items can be recommended and displayed according to business requirements.

[0026] When the user scans the code and adds to the cart, query the partner threshold set of the scanned item. Let the similarity matrix (hereinafter referred to as the partner threshold) of item and item be . First, construct the relationship matrix between users and items in the single order. 1 represents purchase, and 0 represents non-purchase. From this matrix, construct the co-occurrence matrix of items. The co-occurrence matrix represents the number of people who like two items at the same time. Calculate the partner threshold , and the calculation formula is: (3); Among them, represents the number of users who purchase item, represents the number of users who purchase item, represents the number of people who like and items at the same time, and Generated based on the user's historical purchase data. Based on the business design, record the items with the highest partner thresholds and save them in a set. It is necessary to calculate the preferred recommended items for this user based on the matrix. Here, calculate the score based on the user's historical data, and the calculation formula is: (4); Among them, represents the user's predicted score for the item. represents the set of items that the user has historically purchased and settled. represents scanning the items with the highest partner thresholds of the items. In this way, we can obtain the partner threshold list (including items and their thresholds) of the items for a single user and display it to the user according to business requirements.

[0027] At the same time, if the user is not interested in this item, they can click the interactive button "Not interested" on the page. This recommendation will delete it and display the next item. And in the set of items that the user has historically purchased, delete this item until the user repurchases and settles this item. It should be noted that this set is only for algorithm calculation and is not a business storage data set.

[0028] Figure 2 is a flowchart. When the user logs in as a member, query and obtain the previous item score set, use indoor Bluetooth positioning to obtain their real-time location. When the staying time in a certain sales area reaches t0, query the set of items in this sales area, and calculate the predicted scores of these items in this sales area for the user through an algorithm, and display the data according to the interface design, that is, recommend the corresponding items, and update the item score set according to the user's staying time. When the user scans the code and adds the item to the cart, based on the historical information of this item, query its partner threshold set and the barcodes of the partner items, and display the data according to the interface design. When the user adds the item to the cart and settles the item, update the partner threshold set.

[0029] Next, in combination with a preferred embodiment, the content involved in the above embodiment will be described.

[0030] The item recommendation method for the intelligent shopping cart provided in this embodiment involves the following steps: 1. Open the intelligent shopping app; 2. The member logs in and uses the vehicle; 3. Normal business logic execution: User A (this user is not logging in for the first time) strolls among various product cabinets. Indoor Bluetooth positioning is used for real-time positioning. When staying in the beverage area and generating an extraction event that exceeds t0 (assuming t0 is set to 15s), start querying the rating set of User A for the products in the beverage area. If there are 5 products in this area, then there are 5 barcodes corresponding to the 5 products. User A has historical rating records for three of the products (let's call them a, b, and c). Then query the users who have also rated these three products in the last week, obtain their rating sets, substitute them into formula (1) for calculation, find the similar User B, and then substitute each value into formula (2) for calculation to obtain the predicted ratings of User A for the remaining products d and e, generating a decision rating set. This rating set includes the existing ratings of a, b, and c and the predicted ratings of e and d. Assuming the order of the 5 ratings is (b > a > e > c > d), at this time, start querying the shopping mall system to check if there are any promotional activities. If there is a promotional activity for product a, then a will be preferentially recommended and displayed on the system interface.

[0031] 4. When each order is completed, a new product matrix is generated and uploaded to the database for calculation. The partner matrix of each product calculated is saved in the management system as digital property. When User A scans the code to add Coca-Cola Classic Original Flavor to the cart, the system starts querying the partner threshold matrix of Coke recorded in the server. According to the business design, assuming n = 5, then the 5 products with the highest partner thresholds will be obtained. The system queries the user's historical purchased product set, and after obtaining the set, calculates a set of predicted ratings of the user for these 5 products through an algorithm. Assuming the highest rating is for Lay's Italian Red Pepper Flavored Potato Chips with a score of 6.4, then this product will be preferentially displayed.

[0032] 5. If the user is not interested in the recommendation of Lay's Italian Red Pepper Flavored Potato Chips, then they can click the interaction button "Not interested". This recommendation will no longer recommend this product and will instead recommend the product with the second-highest rating. Assuming the Snickers Long Bar with a rating of 6.1 ranks second, then this product will be displayed. If the user is still not interested, then the interaction process will be repeated. If the user does not wish to have the recommendation displayed, they can click the close button to directly close the recommendation function page.

[0033] Embodiment 2. This embodiment provides a product recommendation device for an intelligent shopping cart, including: A data acquisition module, used to obtain the user's real-time location in the shopping mall through Bluetooth beacon positioning and obtain the user's hand movement data through a visual detection module; A first data processing module, used to detect whether the user's stay time at the current location exceeds a preset threshold based on the real-time location and hand movement data, and combine the YOLO algorithm to determine whether the user has taken out a product; if both the stay time exceeds the threshold and the extraction action is detected, it is recorded as the effective stay duration; A second data processing module, configured to construct or update a scoring set of a user for a product corresponding to a real-time location based on the effective stay duration, and use the scoring set as an input to calculate user similarity through a collaborative filtering algorithm based on users, so as to obtain a scoring set of unrated products, thereby obtaining a complete scoring set of products within the area for the user; A first recommendation module, configured to generate a recommended list of the top m products with the highest predicted scores within the current area according to the complete scoring set and when the user stay duration reaches a preset trigger condition, and display the real-time recommendation through an interaction interface; A second recommendation module, configured to, when the user scans a code to add a product to the shopping cart, calculate a partner threshold set and an associated product score of the scanned product through a collaborative filtering algorithm of items based on the complete scoring set and pre-acquired historical purchase data, generate a recommended list of associated products, and display the real-time recommendation through an interaction interface.

[0034] For the specific function implementation of each of the above modules, refer to the relevant content in the method of Embodiment 1, which will not be elaborated here.

[0035] Embodiment 3 provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the methods in Embodiment 1 are implemented.

[0036] Embodiment 4 provides a computer device, including: A memory, configured to store computer programs / instructions; A processor, configured to execute the computer programs / instructions to implement the steps of any one of the methods in Embodiment 1.

[0037] Embodiment 5 provides a computer program product, including computer programs / instructions, and when the computer programs / instructions are executed by a processor, the steps of any one of the methods in Embodiment 1 are implemented.

[0038] The above is only a preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and deformations can still be made, and these improvements and deformations should also be regarded as the protection scope of the present invention.

[0039] Those skilled in the art should understand that the embodiments of the present disclosure can be provided as a method, a system or a computer program product. Therefore, the present disclosure can adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present disclosure can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0040] This disclosure is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the disclosure. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0041] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0042] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more of the flows Figure 1 one or more of the flows and / or blocks Figure 1 or means for implementing the functions specified in one or more of the blocks.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure rather than to limit the scope of its protection. Although the present disclosure has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that after reading the present disclosure, various changes, modifications, or equivalent replacements can still be made to the specific implementation manners of the invention. However, these changes, modifications, or equivalent replacements are all within the scope of the claims of the pending disclosure.

Claims

1. A product recommendation method for an intelligent shopping cart, characterized in that, including: Obtaining the real-time location of the user in the shopping mall through Bluetooth beacon positioning, and obtaining the user's hand movement data through a vision detection module; Based on the real-time location and hand movement data, detecting whether the user's stay time at the current location exceeds a preset threshold, and combining a target detection algorithm to determine whether the user takes out a commodity; if both the stay time exceeds the threshold and the taking-out action is detected, it is recorded as the effective stay duration; Based on the effective stay duration, constructing or updating the scoring set of the commodity corresponding to the real-time location by the user, using the scoring set as the input, calculating the user similarity through a user-based collaborative filtering algorithm, obtaining the scoring set of unrated commodities, so as to obtain the complete scoring set of the commodities in the area by the user; According to the complete scoring set and the user's stay duration reaching the preset trigger condition, generating a recommended list of the top m commodities with the highest predicted scores in the current area, and recommending and displaying them in real time through an interactive interface; When the user scans the code to add a commodity to the shopping cart, based on the complete scoring set and the pre-obtained historical purchase data, calculating the partner threshold set and the associated commodity scores of the scanned commodity through the collaborative filtering algorithm of items, generating an associated commodity recommended list, and recommending and displaying it in real time through an interactive interface.

2. The product recommendation method for an intelligent shopping cart according to claim 1, wherein, The calculation formula of the user similarity is as follows: S(u,v)= ; Among them, S(u, v) represents the similarity, represents the commodity, represents the user 's rating set, represents the user 's evaluation set, represents the user's for 's rating, represents the user for 's rating, and is generated by mapping the effective stay duration, represents the user 's average stay duration, ` represents the user 's average stay duration.

3. The product recommendation method for an intelligent shopping cart according to claim 2, wherein, The score of the unrated commodity, the calculation formula is as follows: ； Among them, represents the predicted user 's rating of the product . represents the neighbor set represents the user's rating of the product .

4. The product recommendation method for an intelligent shopping cart according to claim 1, wherein The calculation of the partner threshold set and the associated commodity scores of the scanned commodity through the collaborative filtering algorithm of items includes: Calculate the partner threshold for scanned goods , and the formula is as follows: ； Among them, represents the number of users who purchase items, represents the number of users who purchase items, represents the number of items that are liked by both and items, and is generated based on the user's historical purchase data; Calculating the associated commodity scores through the partner threshold of the commodity, the formula is as follows: ； Among them, represents the predicted score of the user for the commodity, represents the set of commodities that the user has historically purchased and settled, represents scanning the code for the commodity with the highest partner threshold pieces of commodities.

5. The product recommendation method for an intelligent shopping cart according to claim 1, characterized in that, The method further includes: judging whether the user is a first-time visitor to the store, if so, recommending the commodity with the highest score in the current area.

6. The product recommendation method for an intelligent shopping cart according to claim 1, wherein The method further includes: when recommending and displaying in real time through the interactive interface, supporting the feedback of "not interested" through an interactive button to optimize the recommendation result, and when the user clicks "not interested", excluding the corresponding commodity in real time and updating the recommendation list.

7. The product recommendation method for an intelligent shopping cart according to claim 1, characterized in that, The method further includes: when recommending associated commodities, preferentially displaying partner commodities with targeted discounts for the current commodity, and guiding the user to quickly locate the recommended commodities through the shopping cart navigation function.

8. A product recommendation device for an intelligent shopping cart, characterized in that, including: A data acquisition module, configured to obtain the real-time location of the user in the shopping mall through Bluetooth beacon positioning, and obtain the user's hand movement data through a vision detection module; A first data processing module, configured to detect whether the user's stay time at the current location exceeds a preset threshold based on the real-time location and hand movement data, and combine a target detection algorithm to determine whether the user takes out a commodity; if both the stay time exceeds the threshold and the taking-out action is detected, it is recorded as the effective stay duration; A second data processing module, configured to construct or update the scoring set of the commodity corresponding to the real-time location by the user based on the effective stay duration, use the scoring set as the input, calculate the user similarity through a user-based collaborative filtering algorithm, obtain the scoring set of unrated commodities, so as to obtain the complete scoring set of the commodities in the area by the user; A first recommendation module, configured to generate a recommended list of the top m commodities with the highest predicted scores in the current area according to the complete scoring set and the user's stay duration reaching the preset trigger condition, and recommend and display them in real time through an interactive interface; A second recommendation module, which is used to calculate a set of partner thresholds and associated product scores for a scanned product through a collaborative filtering algorithm for items based on the complete score set and pre-acquired historical purchase data when a user scans a code to add a product to the shopping cart, generate a recommended list of associated products, and recommend and display them in real time through an interactive interface.

9. An electronic device, characterized in that, It includes: A memory for storing computer programs / instructions; A processor for executing the computer programs / instructions to implement the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, it implements the steps of the method according to any one of claims 1-7.

Images