Intelligent retail recommendation method and system based on AI

Abstract

The invention discloses an AI-based intelligent retail recommendation method and system, and the method comprises the steps: obtaining an offline shopping video of a consumer in a supermarket region through a camera device, carrying out the matching of the online account data of the consumer in an account database according to the offline shopping video, and carrying out the matching of the online account data of the consumer in an account database; dividing different types of areas according to different commodity placement positions in the supermarket, identifying effective eye data of the consumer according to the offline shopping video, and combining the effective eye data with a shopping record to analyze potential commodities of the consumer. According to the recommendation weight of the potential commodity, the online commodity list sequence of the consumer is adjusted, the bound commodity of the potential commodity is analyzed in combination with the online account data, and the bound coupon is issued at the online account of the consumer according to the price of the potential commodity and the bound commodity. The method has the characteristics of improving recommendation accuracy and optimizing user experience.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent retail recommendation, and specifically provides an AI-based intelligent retail recommendation method and system. Background Art

[0002] With the rapid development and popularization of Internet technology, e-commerce platforms have emerged in large numbers, seizing the market share of the traditional retail industry at an unprecedented speed and scale. Many physical retailers have seen a decline in customer traffic, sales, and even face the risk of closure due to their inability to compete with e-commerce platforms in terms of price, convenience, and personalized services. To address this challenge, many traditional retailers have started to transform and attempt to integrate online and offline operations to provide a seamless shopping experience. However, many existing retail systems lack the integration of consumers' offline and online shopping behaviors, resulting in a waste of resources for offline and online shopping behaviors and missing out on better recommendation and marketing methods. In addition, when consumers browse online products, the product links uniformly use static pictures or pre-shot retouched videos, and customers cannot understand the authenticity of the products, reducing user trust and user experience. Therefore, it is necessary to design an AI-based intelligent retail recommendation method and system that can improve recommendation accuracy and optimize the user experience. Summary of the Invention

[0003] The purpose of the present invention is to provide an AI-based intelligent retail recommendation method and system to solve the problems raised in the above background art.

[0004] To solve the above technical problems, the present invention provides the following technical solution: An AI-based intelligent retail recommendation method, and the running steps of the method include:

[0005] Step S1: Obtain the offline shopping video of consumers in the supermarket area through a camera device, and match the online account data of the consumers in the account database according to the offline shopping video;

[0006] Step S2: Divide different types of areas according to the placement positions of different products in the supermarket, identify the effective eye data of the consumers according to the offline shopping video, and combine the effective eye data with the shopping records to analyze the potential products of the consumers. The potential products refer to the products that the consumers show a high interest in but have not purchased in the supermarket;

[0007] Step S3: Adjust the order of the consumers' online product lists according to the recommendation weights of the potential products, and analyze the bundled products of the potential products in combination with the online account data. Issue bundled coupons in the consumers' online accounts according to the prices of the potential products and the bundled products. The bundled products refer to the products in the supermarket that are allowed to be bundled and sold with other products.

[0008] wherein, step S2 includes the following steps:

[0009] Step S21: Construct a three-dimensional grid commodity model within the supermarket area. In the three-dimensional grid commodity model, decompose the physical space of the supermarket into three-dimensional grids that conform to the spatial dimensions, and associate corresponding commodity data with each three-dimensional grid;

[0010] Step S22: Collect the facial video stream of the consumer according to the offline shopping video, and identify the eye data of the visual vector in the three-dimensional grid commodity model;

[0011] Step S23: When the fixation duration of a single three-dimensional grid is greater than the threshold according to the temporal continuity of the eye data, determine the eye data as the effective eye data. Construct a commodity interest value analysis model according to the effective eye data, and input the eye data into the commodity interest analysis model to output the interest value of the consumer for the commodity;

[0012] Step S24: Sort the interest values in descending order and associate the corresponding commodity data to establish shopping behavior data, and combine the shopping behavior data with the shopping record to identify the potential commodities of the consumer. wherein, step S24 includes the following steps:

[0013] Step SA1: Construct a periodic purchase feature vector of the consumer according to the historical shopping record;

[0014] Step SA2: Input the periodic purchase feature vector of the commodity in the current shopping record into a regression probability prediction model to obtain the repurchase probability of the commodity in the current shopping record;

[0015] Step SA3: Predict the next repurchase time period of the consumer according to the periodic purchase feature vector. The commodity at the repurchase time period is regarded as a potential commodity, and convert the repurchase probability into a recommendation weight in the online commodity list.

[0016] wherein, step S24 further includes the following steps:

[0017] Step SB1: Compare the shopping behavior data with the current shopping record, and extract the commodities not purchased in the current shopping record from the shopping behavior data. The shopping record includes the current shopping record and the historical shopping record;

[0018] Step SB2: Combine the interest values of the commodities not purchased in the current shopping with the shopping record to construct a next purchase probability function;

[0019] Step SB3: Input the interest value of the goods not purchased this time into the next purchase probability function to obtain the next purchase probability of the consumer for the goods not purchased this time. Consider the goods with the next purchase probability greater than the first threshold as potential goods, and convert the next purchase probability of the potential goods into the recommendation weight for the online product list within the set period for the consumer.

[0020] For step S3, it includes the following steps:

[0021] Step SC1: When the consumer enters the online shopping platform of the supermarket, match the bundling library with the potential goods, compare the recommendation weight with the second threshold, bundle the potential goods with a recommendation weight greater than the second threshold and the potential goods with a recommendation weight less than the second threshold. The bundling library refers to the goods that are allowed to be bundled for preferential promotion.

[0022] Step SC2: Analyze the similarity between the goods in the bundling library and the consumer's shopping records, and consider the goods with a similarity pair threshold higher than the threshold as the bundled goods for the consumer.

[0023] Step SC3: Compare the recommendation weight of the bundlable potential goods with the third threshold. For the bundlable potential goods with a recommendation weight greater than the third threshold, bundle the bundlable potential goods with the bundled goods, where the third threshold is greater than the second threshold.

[0024] For step S3, it further includes the following steps:

[0025] Step SD1: According to the potential goods and the bundled goods, let the consumer make self - matching purchases of the potential goods and the bundled goods. Calculate the face value of the bundled coupon issued to the consumer based on the price face value of the potential goods and the bundled goods matched by the consumer and the preferential intensity allowed in the bundling library.

[0026] For this, the system includes a data acquisition module and a potential goods analysis module:

[0027] The data acquisition module is used to obtain the offline shopping video of consumers in the supermarket area through a camera device, and match the online account data of the consumers in the account database according to the offline shopping video.

[0028] The potential goods analysis module is used to divide different types of areas according to the placement positions of different goods in the supermarket, identify the effective eye movement data of the consumer according to the offline shopping video, and analyze the potential goods of the consumer by combining the effective eye movement data with the shopping records.

[0029] For this, the system further includes a bundled goods recommendation module:

[0030] The bundled product recommendation module is used to adjust the order of the consumer's online product list according to the recommendation weight of the potential product, analyze the bundled products of the potential product in combination with the online account data, and issue a bundled coupon in the consumer's online account according to the prices of the potential product and the bundled products.

[0031] In the third aspect of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the electronic device implements the method described in the first aspect of the present application.

[0032] In the fourth aspect of the present application, a computer-readable storage medium is provided. The computer-readable storage medium is used to store a computer program. When the computer program runs on a computer, the computer executes the method described in the first aspect of the present application.

[0033] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: Through the analysis of effective eye data and shopping behaviors in offline shopping videos, combined with online shopping records, the present invention accurately identifies the degree of consumer interest in products, discovers potential products that consumers have not purchased but have high interest in, optimizes the consumer's online product list based on the potential products, and can push the products that consumers are truly interested in to more prominent positions, improving shopping efficiency and user experience. At the same time, by analyzing the bundled products of potential products and issuing coupons, the purchasing behavior of consumers is guided, and the purchasing intention of consumers and the sales conversion rate of the platform are enhanced. By combining offline shopping behaviors with the online shopping platform, the limitations of the traditional shopping mode are broken, and seamless docking and collaborative optimization of online and offline data are achieved. By guiding online consumption through offline behavior analysis, a more coherent and personalized shopping experience is provided for consumers. Through the accurate identification of potential products and the recommendation of bundled products, merchants can better adjust inventory strategies and reasonably arrange product display and replenishment plans. At the same time, the prediction based on the consumer purchase cycle can help merchants plan promotional activities in advance, reduce inventory backlogs and out-of-stock phenomena, improve the overall efficiency of the supply chain, and thus improve recommendation accuracy and optimize user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The drawings are used to provide an understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0035] Figure 1 It is a schematic flowchart diagram of an AI-based intelligent retail recommendation method provided in Embodiment 1 of the present invention.

[0036] Figure 2 It is a schematic diagram of the module composition of an AI-based intelligent retail recommendation system provided in the second embodiment of the present invention.

[0037] Figure 3 It is a schematic diagram of an electronic device according to an embodiment of the present application. Specific embodiments

[0038] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0039] This embodiment can be applied to the scenario of offline and online recommendation of retail goods. This method can be executed by an AI-based intelligent retail recommendation system provided in this embodiment. Figure 1 It is a schematic flowchart of an AI-based intelligent retail recommendation method provided in the first embodiment of the present invention. This method specifically includes the following steps:

[0040] Step S1: Obtain the offline shopping video of consumers in the supermarket area through a camera device, and match the online account data of the consumers in the account database according to the offline shopping video;

[0041] Step S2: Divide different types of areas according to the placement positions of different goods in the supermarket, identify the effective eye data of the consumers according to the offline shopping video, and combine the effective eye data with the shopping records to analyze the potential goods of the consumers. The potential goods refer to the goods that the consumers show relatively high interest in the supermarket but have not purchased yet;

[0042] Step S3: Adjust the order of the online product list of the consumers according to the recommendation weights of the potential goods, and analyze the bundled goods of the potential goods in combination with the online account data. Issue bundled coupons in the online account of the consumers according to the prices of the potential goods and the bundled goods. The bundled goods refer to the goods in the supermarket that are allowed to be bundled and sold with other goods.

[0043] Specifically, the offline shopping video of consumers is obtained through a camera device and matched with their online account data, so as to achieve seamless connection of online and offline data. For example, the system can identify the stay time or eye fixation of a consumer on a certain product in the supermarket, combine it with the online shopping record, and analyze potential products that the consumer is interested in but has not purchased yet. Then, the system optimizes the consumer's online product list based on the potential products and mines bundled products related to the potential products in combination with the online account data. These bundled products are usually products with potential value-added space but not best-selling products. Through reasonable matching, the purchase intention of consumers can be improved. Finally, the system issues personalized bundled coupons according to the prices of the potential products and the bundled products and the purchase habits of consumers. For example, if a consumer shows interest in a healthy snack in the supermarket but does not purchase it, the system will combine its online purchase record, recommend a healthy drink to match it, and issue a coupon, thus stimulating the consumer's purchase behavior. Through this online-offline integration analysis and recommendation mechanism, accurate insight into consumer behavior and personalized services are achieved, improving the shopping experience and satisfaction of consumers. At the same time, it also provides strong support for merchants to optimize marketing strategies and increase sales conversion rates.

[0044] In some preferred embodiments, the step S2 includes the following steps:

[0045] Step S21: Construct a three-dimensional grid commodity model in the supermarket area. In the three-dimensional grid commodity model, the physical space of the supermarket is decomposed into three-dimensional grids that conform to the space size according to the storage positions and space sizes of different commodities. Each three-dimensional grid is associated with corresponding commodity data, and the commodity data includes: commodity classification label, commodity price, commodity status;

[0046] Step S22: Collect the facial video stream of the consumer according to the offline shopping video, reconstruct the three-dimensional head pose based on the NeRF model, combine iris texture recognition to obtain the visual vector of the consumer, and identify the eye gaze data of the visual vector in the three-dimensional grid commodity model. The eye gaze data includes: the fixation duration t of a single three-dimensional grid, the number of repeated gazes n, and the number of turning points k of the gaze path;

[0047] Step S23: When the fixation duration of a single three-dimensional grid is greater than the threshold according to the time continuity of the eye gaze data, determine the eye gaze data as the effective eye gaze data, and construct the commodity interest value analysis model according to the effective eye gaze data: In the formula, α, β, and γ represent weight coefficients, and the eye gaze data is input into the commodity interest analysis model to output the interest value of the consumer in the commodity;

[0048] Step S24: Sort the interest values in descending order and associate the corresponding product data to establish shopping behavior data, and combine the shopping behavior data with the shopping records to identify potential products for the consumer.

[0049] Specifically, by constructing a three-dimensional grid product model and combining the eye data of consumers, the accurate identification of consumers' interested products and the personalized optimization of the shopping experience are realized. In a supermarket, through the three-dimensional grid product model, the data of each product, including classification labels, prices, and status, can be accurately located and associated. At the same time, by using the NeRF model and iris texture recognition technology to obtain the visual vectors of consumers, the attention of consumers to different products can be accurately captured, such as the fixation duration, the number of repeated gazes, and the number of turning points of the visual path. Based on the product interest value analysis model constructed from these data, the degree of consumers' interest in products can be quantified, and by sorting in descending order and associating product data, shopping behavior data can be established. Combining the shopping behavior data with the shopping records can accurately identify potential products for consumers, that is, those products that consumers show a high interest in but have not purchased yet. It provides data support for the supermarket, helps to optimize product display, inventory management, and marketing strategies, thereby improving operational efficiency and sales performance.

[0050] In some preferred embodiments, step S24 includes the following steps:

[0051] Step SA1: Construct the periodic purchase feature vector C of the consumer according to the historical shopping records i =[a i , b i , c i , d i , where a i represents the average historical purchase interval of the consumer for product i, b i represents the standard deviation of the purchase interval of the consumer for product i, c i represents the purchase frequency of the consumer for product i in the past three months, and d i represents the time difference between the current purchase time of the consumer for product i and the prediction reference date;

[0052] Step SA2: Input the periodic purchase feature vectors of the products in the current shopping record into the regression probability prediction model to obtain the repurchase probability of the products in the current shopping record:

[0053]

[0054] In the formula, ω1 and ω2 represent weight coefficients, and S i represents the promotion gain when product i is in a promotional state;

[0055] Step SA3: Predict the next repurchase time period of the consumer based on the periodic purchase feature vector. The product at the repurchase time period is regarded as a potential product, and the repurchase probability is converted into a recommendation weight in the online product list.

[0056] Specifically, by analyzing the periodic purchase feature vector of the consumer and applying a regression probability prediction model, accurate prediction of the consumer's repurchase behavior and personalized recommendation are achieved. Based on the periodic purchase feature vector, the next repurchase time period of the consumer is predicted, and the repurchase probability is converted into a recommendation weight in the online product list, thereby realizing accurate recommendation. This approach not only improves the shopping experience of consumers, enabling them to receive recommendations for products they are interested in in a timely manner, but also helps the supermarket optimize inventory management and marketing strategies, improving operational efficiency and sales performance.

[0057] In some preferred embodiments, step S24 further includes the following steps:

[0058] Step SB1: Compare the shopping behavior data with the current shopping record, and extract the products not purchased in the current shopping behavior data. The shopping record includes the current shopping record and historical shopping records;

[0059] Step SB2: Combine the interest values of the products not purchased in the current shopping with the shopping record to construct a next purchase probability function:

[0060]

[0061] In the formula, Q avg represents the average interest in expected purchase in the later stage of the consumer's historical interest value, R i represents the feature deviation between product i and the shopping record, T predicted represents the predicted purchase cycle, t current represents the current time, Δt represents the interest decay time window, and σ represents the degree of dispersion measuring the time distribution;

[0062] Step SB3: Input the interest values of the products not purchased in the current shopping into the next purchase probability function to obtain the next purchase probability of the consumer for the products not purchased in the current shopping. The products with the next shopping probability greater than the first threshold are regarded as potential products, and the next purchase probability of the potential products is converted into a recommendation weight in the online product list for the consumer within the set cycle.

[0063] Specifically, by comparing the shopping behavior data with the current shopping record, constructing a next purchase probability function in combination with the interest value of the goods not purchased this time, and inputting the interest value into this function to calculate the next purchase probability, the accurate prediction and personalized recommendation of consumers' potential purchase behavior are realized. For example, assume that a certain consumer shows a relatively high interest in a certain brand of shampoo in the supermarket (such as long-time gazing, picking up the goods multiple times, etc.), but finally does not purchase. The system will combine the interest value of this shampoo with the consumer's shopping record, consider factors such as the consumer's historical interest average, the deviation of product characteristics, and the predicted purchase cycle, and construct a next purchase probability function. If the calculated next purchase probability is greater than the first threshold (such as 0.7), then this shampoo is regarded as a potential product, and its next purchase probability is converted into the recommendation weight of the online product list.

[0064] In some optional embodiments, the step SB2 includes the following steps:

[0065] Step SB21: Extract the product characteristics in the shopping record to construct the shopping feature vector of the consumer, and extract the same-dimensional feature vector of the goods not purchased this time to calculate the matching feature value R between it and the shopping feature vector;

[0066] Step SB22: Combine the matching feature value R with the correction influence of time on price to obtain the feature deviation of the goods not purchased this time In the formula, α represents the trend attenuation coefficient, R t represents the feature deviation of product i at the current time point, R t-1 represents the feature deviation of product i at the previous time point, represents the time correction coefficient, and q represents the historical average purchase cycle of the consumer.

[0067] Specifically, by extracting the product characteristics in the shopping record to construct the shopping feature vector of the consumer, and calculating the matching feature value R between the same-dimensional feature vector of the goods not purchased this time and the shopping feature vector. For example, assume that the consumer's shopping record shows that they often buy organic vegetables and low-sugar foods, then the shopping feature vector will reflect these preference characteristics. For a certain type of organic vegetables not purchased this time, the matching feature value R between its feature vector and the consumer's shopping feature vector is relatively high, indicating that this product is more in line with the consumer's preferences. Combine the matching feature value R with the correction influence of time on price to obtain the feature deviation of the goods not purchased this time. For example, if the consumer's historical average purchase cycle is two weeks, and the current time point is more than two weeks away from the last purchase, and the price of the product has fluctuated, then the feature deviation of the product can be calculated through calculation. This deviation can reflect the change in the attractiveness of the product to the consumer at the current time point. Through the above steps, the accurate analysis and prediction of consumers' purchase behavior are realized.

[0068] In some preferred embodiments, step S3 includes the following steps:

[0069] Step SC1: When the consumer enters the online shopping platform of the supermarket, match the bundling library with the potential products, extract the bundlable potential products and the recommendation weights, compare the recommendation weights with a second threshold, and bundle the potential products with recommendation weights greater than the second threshold with those less than the second threshold. The bundling library refers to products that are allowed to be bundled for preferential promotions.

[0070] Step SC2: Analyze the similarity between the products in the bundling library and the consumer's shopping records, and bundle the products with a similarity pair threshold higher than the threshold for the consumer.

[0071] Step SC3: Compare the recommendation weights of the bundlable potential products with a third threshold, and bundle the bundlable potential products with recommendation weights greater than the third threshold with the bundled products, where the third threshold is greater than the second threshold.

[0072] Specifically, accurate analysis of consumers' purchase behaviors and personalized bundling recommendations are achieved. For example, when a consumer enters the online shopping platform of the supermarket, the system matches the bundling library with the potential products to extract the bundlable potential products and their recommendation weights. Suppose the consumer often buys milk and bread. Through step SC1, the system bundles and recommends potential products with recommendation weights higher than the second threshold (such as milk) with those with recommendation weights lower than the second threshold (such as bread) to form a preferential combination of milk + bread. In step SC2, the system analyzes the similarity between the products in the bundling library and the consumer's shopping records, and uses the products with a similarity higher than the threshold (such as a certain brand of milk previously purchased by the consumer) as the bundled products. In step SC3, the system compares the recommendation weights of the bundlable potential products with the third threshold. For products with recommendation weights greater than the third threshold (such as a new brand of milk), it bundles and recommends them with the bundled products (such as bread). Through this multi-dimensional analysis and recommendation strategy, the supermarket can provide more attractive preferential combinations for consumers, increase consumers' purchase willingness and purchase volume, and at the same time improve the sales efficiency of products and the inventory turnover rate.

[0073] In some preferred embodiments, step S3 further includes the following steps:

[0074] Step SD1: Based on the potential products and the bundled products, let the consumer make self-matched purchases of the potential products and the bundled products, and calculate the face value of the bundled coupons issued to the consumer according to the price face values of the potential products and the bundled products matched by the consumer and the preferential intensity allowed in the bundling library.

[0075] Specifically, it realizes the dynamic calculation and issuance of personalized bundled coupon face values according to consumers' self-matching purchase behaviors of potential products and bundled products. For example, assume that a consumer self-matches and purchases a potential product (such as a newly launched healthy snack) and a bundled product (such as frequently purchased milk) on a shopping platform. The system will calculate based on the price face values of these two products. If the set preferential rule is to issue coupons at 5% of the total product price, then the consumer will receive the corresponding bundled coupon, allowing the consumer to pick up the order online or consume offline. In this way, the system can provide accurate and personalized coupons according to consumers' actual purchase behaviors and preferences, enhancing consumers' stickiness to the shopping platform and purchase willingness. At the same time, this dynamic coupon issuance mechanism also improves the marketing efficiency of merchants, reduces ineffective promotions, and enhances the overall operational efficiency.

[0076] Based on the same inventive concept as the above method embodiment, the embodiment of the present invention also provides an AI-based intelligent retail recommendation system. Figure 2 It is a schematic diagram of the module composition of an AI-based intelligent retail recommendation system provided by the embodiment of the present invention, as Figure 2 shown. The system includes a data collection module, a potential product analysis module, and a bundled product recommendation module:

[0077] The data collection module is used to obtain the offline shopping videos of consumers in the supermarket area through a camera device, and match the online account data of the consumers in the account database according to the offline shopping videos;

[0078] The potential product analysis module is used to divide different types of areas according to the placement positions of different products in the supermarket, identify the effective eye movement data of the consumers according to the offline shopping videos, and analyze the potential products of the consumers by combining the effective eye movement data with the shopping records;

[0079] The bundled product recommendation module is used to adjust the order of the consumers' online product lists according to the recommendation weights of the potential products, analyze the bundled products of the potential products in combination with the online account data, and issue bundled coupons in the consumers' online accounts according to the prices of the potential products and the bundled products.

[0080] It should be noted that although several units or subunits of the device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present application, the features and functions of the two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be divided and embodied by multiple units.

[0081] Based on the same inventive concept as the above method embodiments, an electronic device is also provided in an embodiment of the present application. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the electronic device implements the control method in the above embodiments.

[0082] In one embodiment, the electronic device may be a server. In this embodiment, the structure of the electronic device may be as Figure 3 shown, including a memory 2001, a communication module 2003, and one or more processors 2002.

[0083] The memory 2001 is used to store the computer program executed by the processor 2002. The memory 2001 may mainly include a program storage area and a data storage area. Among them, the program storage area may store an operating system and programs required to run the instant messaging function, etc.; the data storage area may store various instant messaging information and operation instruction sets, etc.

[0084] The memory 2001 may be a volatile memory, such as a random-access memory (RAM); the memory 2001 may also be a non-volatile memory, such as a read-only memory, a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD); or the memory 2001 is any other medium that can be used to carry or store a desired computer program in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2001 may be a combination of the above memories.

[0085] The processor 2002 may include one or more central processing units (CPUs) or be a digital processing unit, etc. The processor 2002 is used to implement the above audio data processing method when calling the computer program stored in the memory 2001.

[0086] The communication module 2003 is used to communicate with terminal devices and other servers.

[0087] In the embodiments of the present application, the specific connection medium between the above memory 2001, communication module 2003, and processor 2002 is not limited. In the embodiments of the present application Figure 3 it is connected between the memory 2001 and the processor 2002 through a bus 2004, and the bus 2004 is in Figure 3It is described by an arrow in the figure. The connection manners between other components are only for illustrative purposes and are not limited thereto. The bus 2004 may be divided into an address bus, a data bus, a control bus, etc. For ease of description, Figure 3 only one arrow is used for description in the figure, but it does not describe that there is only one bus or one type of bus.

[0088] Based on the same inventive concept as the above method embodiment, an embodiment of the present invention further provides a computer-readable storage medium, which is used to store a computer program. When the computer program runs on a computer, the electronic device is enabled to implement the control method in the above embodiment. The computer-readable storage medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the readable storage medium include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0089] Based on the same inventive concept as the above method embodiment, an embodiment of the present invention further provides a computer program product, which includes a computer program. When the program product runs on an electronic device, the computer program is used to enable the electronic device to execute the steps in the control method according to various exemplary embodiments described above in this specification. The program product may adopt any combination of one or more readable media. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate a device for realizing the functions specified in Figure 1 one process or multiple processes and / or Figure 1 one block or multiple blocks.

[0090] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they know the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments and all changes and modifications falling within the scope of the present application.

Claims

1. An AI-based intelligent retail recommendation method, characterized in that: Obtain the offline shopping video of consumers in the supermarket area through a camera device, and match the online account data of the consumers in the account database according to the offline shopping video; Divide different types of areas according to the placement positions of different commodities in the supermarket, identify the effective eye data of the consumers according to the offline shopping video, and combine the effective eye data with the shopping records to analyze the potential commodities of the consumers. The potential commodities refer to the commodities that the consumers show relatively high interest in but have not purchased in the supermarket; Adjust the order of the online commodity list of the consumers according to the recommendation weights of the potential commodities, combine the online account data to analyze the bundled commodities of the potential commodities, and issue bundled coupons in the online accounts of the consumers according to the prices of the potential commodities and the bundled commodities. The bundled commodities refer to the commodities in the supermarket that are allowed to be bundled and sold with other commodities.

2. The intelligent retail recommendation method based on AI according to claim 1, wherein: The step of dividing different types of areas according to the placement positions of different commodities in the supermarket, identifying the effective eye data of the consumers according to the offline shopping video, and combining the effective eye data with the shopping records to analyze the potential commodities of the consumers includes: Construct a three-dimensional grid commodity model in the supermarket area. In the three-dimensional grid commodity model, decompose the physical space of the supermarket into three-dimensional grids that conform to the space size, and each three-dimensional grid is associated with corresponding commodity data; Collect the facial video stream of the consumers according to the offline shopping video, and identify the eye data of the visual vector in the three-dimensional grid commodity model; When the fixation duration of a single three-dimensional grid is greater than the threshold according to the time continuity of the eye data, determine that the eye data is the effective eye data. Construct a commodity interest value analysis model according to the effective eye data, and input the eye data into the commodity interest analysis model to output the interest value of the consumers for the commodities; Sort the interest values in descending order and associate the corresponding commodity data to establish shopping behavior data, and combine the shopping behavior data with the shopping records to identify the potential commodities of the consumers.

3. The intelligent retail recommendation method based on AI according to claim 2, characterized in that: The step of sorting the interest values in descending order and associating the corresponding commodity data to establish shopping behavior data, and combining the shopping behavior data with the shopping records to identify the potential commodities of the consumers includes: Construct a periodic purchase feature vector of the consumers according to the historical shopping records; Input the periodic purchase feature vector of the commodities in the current shopping record into a regression probability prediction model to obtain the repurchase probability of the commodities in the current shopping record; Predict the next repurchase time period of the consumers according to the periodic purchase feature vector, regard the commodities at the repurchase time period as potential commodities, and convert the repurchase probability into a recommendation weight in the online commodity list.

4. The intelligent retail recommendation method based on AI according to claim 3, wherein: The step of sorting the interest values in descending order and associating the corresponding commodity data to establish shopping behavior data, and combining the shopping behavior data with the shopping records to identify the potential commodities of the consumers further includes: Compare the shopping behavior data with the current shopping record, and extract the goods not purchased in the current shopping from the shopping behavior data. The shopping record includes the current shopping record and historical shopping records; Combine the interest values of the goods not purchased in the current shopping with the shopping record to construct a next purchase probability function; Input the interest values of the goods not purchased in the current shopping into the next purchase probability function to obtain the next purchase probability of the consumer for the goods not purchased in the current shopping. Regard the goods with the next purchase probability greater than the first threshold as potential goods, and convert the next purchase probability of the potential goods into the recommendation weight for the online product list of the consumer within the set period.

5. The AI-based intelligent retail recommendation method according to claim 4, wherein: Adjust the order of the consumer's online product list according to the recommendation weight of the potential goods, analyze the bundled goods of the potential goods in combination with the online account data, and issue bundled coupons in the consumer's online account according to the prices of the potential goods and the bundled goods, including: When the consumer enters the online shopping platform of the supermarket, match the bundling library with the potential goods, compare the recommendation weight with the second threshold, and bundle the potential goods greater than the second threshold with the potential goods less than the second threshold. The bundling library refers to the goods that are allowed to be bundled for preferential promotion; Analyze the similarity between the goods in the bundling library and the consumer's shopping record, and the goods with a similarity pair threshold higher than the threshold are the bundled goods for the consumer; Compare the recommendation weight of the bundlable potential goods with the third threshold. For the bundlable potential goods with the recommendation weight greater than the third threshold, bundle the bundlable potential goods with the bundled goods, where the third threshold is greater than the second threshold.

6. The intelligent retail recommendation method based on AI according to claim 5, characterized in that: Adjust the order of the consumer's online product list according to the recommendation weight of the potential goods, analyze the bundled goods of the potential goods in combination with the online account data, and issue bundled coupons in the consumer's online account according to the prices of the potential goods and the bundled goods, further including: Run the consumer to purchase the potential goods and the bundled goods by themselves according to the potential goods and the bundled goods. Calculate the face value of the bundled coupons issued to the consumer according to the face value of the potential goods and the bundled goods matched by the consumer and the preferential intensity allowed in the bundling library.

7. An AI-based intelligent retail recommendation system, characterized in that: The system includes a data collection module and a potential goods analysis module: The data collection module is used to obtain the offline shopping video of consumers in the supermarket area through a camera device, and match the online account data of the consumers in the account database according to the offline shopping video; The potential goods analysis module is used to divide different types of areas according to the placement positions of different goods in the supermarket, identify the effective eye movement data of the consumers according to the offline shopping video, and analyze the potential goods of the consumers by combining the effective eye movement data with the shopping record.

8. An AI-based intelligent retail recommendation system according to claim 7, characterized in that: The system further includes a bundled goods recommendation module: The bundled product recommendation module is used to adjust the order of the consumer's online product list according to the recommendation weight of the potential product, analyze the bundled products of the potential product in combination with the online account data, and issue a bundled coupon in the consumer's online account according to the prices of the potential product and the bundled products.

9. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, the electronic device implements the method according to any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program, and when the computer program runs on a computer, the computer executes the method according to any one of claims 1 to 6.

Images