Vending machine commodity purchase recommendation method and device based on multiple factors and dynamic classification

Abstract

The invention discloses a vending machine commodity purchase recommendation method and device based on multiple factors and dynamic classification, and the method comprises the steps: constructing a recommendation factor priority queue according to the scene difference between a mobile terminal and a machine terminal, and generating an initial candidate commodity pool through a multi-level dynamic filling mechanism; manual configuration, user behaviors and complementary commodity factors are preferentially called, and insufficient parts are complemented by urban hot-sell commodities. And low-value and sensitive commodities are eliminated by adopting a price threshold value, a classification blacklist and a state filtering rule, and first-level classification commodities are forcibly complemented through a reverse-order backtracking strategy, so that category balance of a recommendation list is ensured. And furthermore, the Hash duplicate removal and interface-level preferential isolation technologies are combined, so that commodity repeated recommendation and preferential rule conflicts are eliminated. According to the invention, through a multi-factor collaborative recommendation and dynamic classification compensation mechanism, the proportion of the unsalable goods is significantly reduced, the recommendation accuracy of the high-value goods is improved, and the diversity and consumption experience of the goods at the user side are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of vending machine management, and in particular to a method and device for recommending additional purchases of vending machine products based on multiple factors and dynamic classification. Background Art

[0002] Vending machine product recommendation technology is primarily driven by static rules based on a single dimension, such as generating recommendation lists based on weather conditions, inventory status, or basic user behavior data. Typical solutions include matching products based on environmental parameters such as temperature and humidity, prioritizing overstocked products to reduce the risk of unsaleable items, or recommending similar categories based on historical purchase records. This type of technology generally uses linear recommendation logic, relies on manually preset weights (such as a fixed recommendation weight of 50% of inventory), lacks a dynamic coordination mechanism, and the data source is limited to a single dimension. It fails to achieve a deep integration of multiple factors such as user behavior, product attributes, and real-time sales, resulting in a limited match between recommendation results and market dynamics and user needs.

[0003] Currently, existing solutions, due to technical limitations, lead to multiple operational and user experience conflicts. Under a static rule-based system, recommendation priorities cannot dynamically adjust based on real-time user behavior, resulting in under-exposure of high-value products. Data silos also lead to fragmented analysis across dimensions like weather, inventory, and user behavior. For example, high temperatures are not integrated with user preferences to uncover potential demand. While inventory-driven recommendations reduce the risk of unsold items, they exacerbate the accumulation of low-profitable items. Furthermore, the bias rate in first-level categories is high (e.g., over-exposure of snack foods and chronic omission of healthy foods). Furthermore, the lack of mutually exclusive discount rules results in added-to-cart items participating in discount promotions, triggering user complaints. The filtering of sensitive products relies on manual inspections, resulting in response delays exceeding 24 hours. The daily update mechanism for sales statistics fails to capture sudden changes in demand during promotional periods, further weakening the timeliness and accuracy of recommendations and creating a vicious cycle of "data lag, recommendation bias, and operational losses."

[0004] Therefore, how to invent and develop a vending machine product purchase recommendation method based on multiple factors and dynamic classification, which can solve the problems of the existing technology such as the high proportion of slow-moving products, single recommendation factors, uneven classification coverage and poor experience, has become an urgent problem to be solved. Summary of the Invention

[0005] To this end, the present invention provides a method and device for recommending add-on items for vending machines based on multiple factors and dynamic classification. By optimizing the recommendation logic, this method reduces the exposure of slow-moving items, improves the accuracy of high-value product recommendations for vending machines, and ensures balanced category coverage. This method addresses existing issues such as a high proportion of slow-moving items, a single recommendation factor, uneven category coverage, and a poor user experience.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for recommending additional items for vending machines based on multiple factors and dynamic classification, comprising:

[0007] Based on the differences in scenarios between mobile and machine terminals, a mobile terminal recommendation factor priority queue and a machine terminal recommendation factor priority queue are constructed respectively; recommendation factors are sequentially called based on the mobile terminal recommendation factor priority queue to generate an initial candidate product pool;

[0008] Filtering the initial candidate product pool by setting multi-dimensional filtering rules to obtain a filtered candidate product pool;

[0009] Performing a first-level category quantity check on the filtered candidate product pool; if the first-level category quantity does not meet the minimum quantity requirement, rescreening and supplementing the missing category products in the reverse order of the mobile terminal recommendation factor priority queue to obtain a balanced category product pool;

[0010] The classified balanced product list is subjected to product deduplication and discount isolation processing to obtain a final product recommendation pool.

[0011] As a preferred solution for the method of recommending additional items for vending machines based on multiple factors and dynamic classification, the mobile terminal's recommendation factor priority queue is: manual configuration > user behavior > complementary products > product attributes > city hot sales;

[0012] The machine-side recommendation factor priority queue is manually configured as follows: > inventory turnover rate > complementary products > product attributes > city hot sales.

[0013] As a preferred solution of the method for recommending additional products for vending machines based on multiple factors and dynamic classification, in the process of sequentially calling recommendation factors based on the mobile terminal recommendation factor priority queue to generate the initial candidate product pool:

[0014] The candidate products for the recommendation factor are determined through a comprehensive scoring formula, and each non-terminal recommendation factor only recommends one product each time it is called. Unfilled positions are dynamically supplemented by the city's hot-selling factors.

[0015] The comprehensive scoring formula is:

[0016] S=α·M+β·U+γ·C+δ·A+ε·H

[0017] Where S is the total score; M is the manually configured weight, M=1 for configured products, otherwise M=0; U is the user behavior score; C is the correlation between complementary products; A is the product attribute score; H is the city's hot-selling coefficient; α, β, γ, δ, and ε are all weight coefficients, and they satisfy α>β>γ>δ>ε.

[0018] As a preferred solution of the method for recommending additional products for vending machines based on multiple factors and dynamic classification, in the process of filtering and screening the initial candidate product pool by setting multi-dimensional filtering rules:

[0019] The multi-dimensional filtering rules set include: goods to be paid, goods with an original price of ≤ 4 yuan, and goods in specific categories.

[0020] As a preferred solution for the vending machine product add-on recommendation method based on multiple factors and dynamic classification, a dynamic hot-selling pool is calculated based on basic sales volume and new product weighting strategy; the dynamic hot-selling pool provides real-time data support for the city hot-selling factors; the calculation formula of the dynamic hot-selling pool is:

[0021]

[0022] Where N is the monthly sales volume; d is the date; μ is the attenuation coefficient, usually 0.05; sale d is the sales volume on the past d day; e -μ·(30-d) is the exponential decay factor;

[0023]

[0024] Where N f is the final sales volume; θ is the weighting coefficient, which is 0.5; t is the number of days on the shelf; T=7 is the new product protection period.

[0025] The present invention also provides a device for recommending additional items for vending machines based on multiple factors and dynamic classification, comprising:

[0026] An initial candidate product pool generation module is configured to construct a mobile-side recommendation factor priority queue and a machine-side recommendation factor priority queue based on the scenario differences between the mobile side and the machine side, respectively; and to sequentially call recommendation factors based on the mobile-side recommendation factor priority queue to generate an initial candidate product pool;

[0027] A filtered candidate commodity pool acquisition module is used to filter and screen the initial candidate commodity pool by setting multi-dimensional filtering rules to obtain a filtered candidate commodity pool;

[0028] A module for obtaining a balanced product list by category is used to verify the number of first-level categories in the filtered candidate product pool; if the number of first-level categories does not meet the minimum quantity requirement, the module re-screens and supplements the missing category products in the reverse order of the mobile terminal recommendation factor priority queue to obtain a balanced product pool by category;

[0029] The final product recommendation pool acquisition module is used to perform product deduplication and discount isolation processing on the classified balanced product list to obtain the final product recommendation pool.

[0030] As a preferred solution of the device for recommending additional products for vending machines based on multiple factors and dynamic classification, in the initial candidate product pool generation module, the mobile terminal's recommendation factor priority queue is: manual configuration > user behavior > complementary products > product attributes > city hot sales;

[0031] The machine-side recommendation factor priority queue is manually configured as follows: > inventory turnover rate > complementary products > product attributes > city hot sales.

[0032] As a preferred solution of the device for recommending additional products for vending machines based on multiple factors and dynamic classification, in the initial candidate product pool generation module, when the recommendation factors are sequentially called based on the mobile terminal recommendation factor priority queue to generate the initial candidate product pool:

[0033] The candidate products for the recommendation factor are determined through a comprehensive scoring formula, and each non-terminal recommendation factor only recommends one product each time it is called. Unfilled positions are dynamically supplemented by the city's hot-selling factors.

[0034] The comprehensive scoring formula is:

[0035] S=α·M+β·U+γ·C+δ·A+ε·H

[0036] Where S is the total score; M is the manually configured weight, M=1 for configured products, otherwise M=0; U is the user behavior score; C is the correlation between complementary products; A is the product attribute score; H is the city's hot-selling coefficient; α, β, γ, δ, and ε are all weight coefficients, and they satisfy α>β>γ>δ>ε.

[0037] As a preferred solution of the device for recommending additional products for vending machines based on multiple factors and dynamic classification, in the module for obtaining the filtered candidate product pool, during the process of filtering the initial candidate product pool using the set multi-dimensional filtering rules:

[0038] The multi-dimensional filtering rules set include: goods to be paid, goods with an original price of ≤ 4 yuan, and goods in specific categories.

[0039] As a preferred solution for the device for recommending additional products for vending machines based on multiple factors and dynamic classification, the initial candidate product pool generation module calculates a dynamic hot-selling pool based on basic sales volume and new product weighting strategy; the dynamic hot-selling pool provides real-time data support for the city hot-selling factor; the calculation formula for the dynamic hot-selling pool is:

[0040]

[0041] Where N is the monthly sales volume; d is the date; μ is the attenuation coefficient, usually 0.05; sale d is the sales volume on the past d day; e -μ·(30-d)is the exponential decay factor;

[0042]

[0043] Where N f is the final sales volume; θ is the weighting coefficient, which is 0.5; t is the number of days on the shelf; T=7 is the new product protection period.

[0044] The present invention has the following advantages: Based on the scenario differences between mobile and machine terminals, the present invention constructs a mobile-side recommendation factor priority queue and a machine-side recommendation factor priority queue, respectively; based on the mobile-side recommendation factor priority queue, recommendation factors are sequentially called to generate an initial candidate product pool; the initial candidate product pool is filtered and screened by setting multi-dimensional filtering rules to obtain a filtered candidate product pool; the filtered candidate product pool is verified for the number of first-level categories; if the number of first-level categories does not meet the minimum quantity requirement, the missing category products are re-screened and supplemented in the reverse order of the mobile-side recommendation factor priority queue to obtain a balanced category product pool; the balanced category product list is deduplicated and preferentially isolated to obtain a final product recommendation pool. The present invention reduces the exposure of slow-moving products, increases the probability of recommending high-value products, and improves recommendation accuracy through multi-factor priority and dynamic filling. The present invention ensures product diversity, avoids duplicate recommendations, and optimizes the user experience through a forced balancing and deduplication mechanism based on category proportions. The present invention reduces losses caused by invalid recommendations and preferential conflicts through blacklist filtering and preferential isolation rules, thereby reducing operating costs. The present invention solves the problems in the prior art such as a high proportion of slow-moving products, a single recommendation factor, uneven classification coverage, and poor user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and those skilled in the art can, without inventive effort, derive other implementation drawings based on the provided drawings.

[0046] The structures, proportions, sizes, etc. illustrated in this specification are intended solely to complement the contents disclosed herein and to facilitate understanding and reading by persons skilled in the art. They are not intended to limit the conditions under which the present invention may be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportions, or adjustments in sizes, without affecting the efficacy and objectives of the present invention, shall remain within the scope of the technical contents disclosed herein.

[0047] Figure 1This is a flow chart of the method for recommending additional items for vending machines based on multiple factors and dynamic classification, provided in Example 1 of the present invention;

[0048] Figure 2 This is a schematic diagram of the process of generating an initial candidate product pool in the method for recommending additional products for vending machines based on multiple factors and dynamic classification provided in Example 1 of the present invention;

[0049] Figure 3 This is a schematic diagram of the first-level classification quantity verification process in the vending machine product add-on recommendation method based on multiple factors and dynamic classification provided in Example 1 of the present invention;

[0050] Figure 4 This is a schematic diagram of the product filtering and product deduplication process in the vending machine product add-on recommendation method based on multiple factors and dynamic classification provided in Example 1 of the present invention;

[0051] Figure 5 Schematic diagram of the architecture of the device for recommending additional products for vending machines based on multiple factors and dynamic classification provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0052] The following describes the implementation of the present invention using specific embodiments. Those skilled in the art will readily understand the other advantages and benefits of the present invention from the disclosure herein. Obviously, the embodiments described are only a portion of the present invention, not all of it. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0053] Example 1

[0054] See also Figure 1 Embodiment 1 of the present invention provides a method for recommending additional items for vending machines based on multiple factors and dynamic classification, comprising the following steps:

[0055] S1. Based on the differences in scenarios between mobile and machine terminals, a mobile terminal recommendation factor priority queue and a machine terminal recommendation factor priority queue are respectively constructed; recommendation factors are sequentially called based on the mobile terminal recommendation factor priority queue to generate an initial candidate product pool;

[0056] S2. Filter the initial candidate product pool by setting multi-dimensional filtering rules to obtain a filtered candidate product pool;

[0057] S3. Verify the number of first-level categories in the filtered candidate product pool; if the number of first-level categories does not meet the minimum quantity requirement, re-screen and supplement the missing category products in the reverse order of the mobile terminal recommendation factor priority queue to obtain a balanced product pool;

[0058] S4. De-duplicate the classified and balanced product list and isolate the discounts to obtain a final product recommendation pool.

[0059] In this embodiment, in step S1, based on the scenario differences between the mobile terminal and the machine terminal, a mobile terminal recommendation factor priority queue and a machine terminal recommendation factor priority queue are respectively constructed; based on the mobile terminal recommendation factor priority queue, recommendation factors are sequentially called to generate an initial candidate product pool;

[0060] Specifically, based on the differences in scenarios between mobile and machine terminals, we construct a mobile terminal recommendation factor priority queue and a machine terminal recommendation factor priority queue respectively;

[0061] The mobile recommendation factor priority ranking is: manual configuration > user behavior > complementary products > product attributes > city hot sellers. User behavior factors are derived by analyzing a user's historical orders, for example, products with a purchase frequency of ≥3 times per month and a click-through rate of >15% for add-to-cart items. Complementary products are identified using the Apriori algorithm to identify co-occurrence patterns among products with a support score >5% and a confidence score >30%.

[0062] The priority queue of machine-side recommendation factors is: manual configuration > inventory turnover rate > complementary products > product attributes > city hot sales;

[0063] The formula for calculating inventory turnover is:

[0064] Turnover rate = (30-day sales) / (current inventory + safety stock)

[0065] Complementary products are identified by matching product attribute tags (such as "coffee + creamer", "instant noodles + ham").

[0066] In this embodiment, the recommendation factors are sequentially called based on the mobile terminal recommendation factor priority queue to generate an initial candidate product pool;

[0067] Specifically, the candidate products for the recommendation factor are determined through a comprehensive scoring formula, and each non-terminal recommendation factor only recommends one product each time it is called, and the unfilled positions are dynamically supplemented by the city's hot-selling factors;

[0068] The comprehensive scoring formula is:

[0069] S=α·M+β·U+γ·C+δ·A+ε·H

[0070] Where S is the total score; M is the manually configured weight, M=1 for configured products, otherwise M=0; U is the user behavior score; C is the correlation between complementary products; A is the product attribute score; H is the city's hot-selling coefficient; α, β, γ, δ, and ε are all weight coefficients, and they satisfy α>β>γ>δ>ε.

[0071] In this embodiment, a dynamic hot-selling pool is calculated based on basic sales volume and a new product weighting strategy. The dynamic hot-selling pool provides real-time data support for the city hot-selling factor. The calculation formula for the dynamic hot-selling pool is:

[0072]

[0073] Where N is the monthly sales volume; d is the date; μ is the attenuation coefficient, usually 0.05; sale d is the sales volume on the past d day; e -μ·(30-d) is the exponential decay factor;

[0074]

[0075] Where N f is the final sales volume; θ is the weighting coefficient, which is 0.5; t is the number of days on the shelf; T=7 is the new product protection period.

[0076] In this embodiment, Figure 2 As shown in the figure, the specific filling mechanism is as follows: the manual configuration layer forcibly recommends two specified products (such as new product trial packs); the user behavior layer only selects the top 1 high-weighted product each time it is called; if there are no available products in the current layer, the next layer is immediately triggered; and finally, the vacancies are filled by the city's hot-selling products in descending order of sales.

[0077] In this embodiment, in step S2, the initial candidate product pool is filtered and screened by setting multi-dimensional filtering rules to obtain a filtered candidate product pool;

[0078] The multi-dimensional filtering rules include: commodities to be paid, commodities with an original price of ≤ 4 yuan, and commodities in specific categories.

[0079] Specifically, such as Figure 4 As shown, price filtering: exclude products with original price ≤ 4 yuan (such as mineral water, paper towels);

[0080] Category filtering: exclude specific categories by GCID prefix;

[0081] Status filtering: Shield items in the user's pending payment order to prevent duplicate recommendations;

[0082] Sensitive word filtering: Automatically block product names containing keywords such as "trial" and "gift".

[0083] In this embodiment, in step S3, the filtered candidate product pool is checked for the number of first-level categories; if the number of first-level categories does not meet the minimum quantity requirement, the missing category products are re-screened and supplemented in the reverse order of the mobile terminal recommendation factor priority queue to obtain a balanced product pool;

[0084] Specifically, such as Figure 3 As shown, the add-to-cart list must contain at least a specified number of specific first-level category products, such as 2 packaged beverages and 2 snack foods; if the requirements are not met, the list will be supplemented in the reverse order of the mobile terminal recommendation factor priority queue.

[0085] The reverse tracing implementation steps are:

[0086] T1. Detect the current classification gap (e.g., there is only one packaged beverage);

[0087] T2, search from the end of the priority queue (the popular city layer) in reverse order;

[0088] The search order is: hot selling in the city → product attributes → complementary products → user behavior;

[0089] T3. Capture products in the target category by sales ranking (prioritize the top 20% in sales, and if that is insufficient, relax to the top 50%).

[0090] In this embodiment, in step S4, the categorized balanced product list is subjected to product deduplication and discount isolation processing to obtain a final product recommendation pool.

[0091] Specifically, such as Figure 4 As shown, the classified and balanced product list will be deduplicated and the preferential treatment will be isolated;

[0092] Product deduplication refers to the strict deduplication of added-to-cart products and between added-to-cart products and main products (except for manually configured products).

[0093] Among them, a hash table is used to record the recommended product IDs to ensure that there are no duplications in the add-to-cart list.

[0094] Discount isolation means that added-to-cart items do not participate in any discount activities to avoid conflicts with the discount rules of the main product.

[0095] Specifically, preferential isolation is achieved through interface-level isolation and amount-level isolation.

[0096] Interface-level isolation: When calling the marketing system to calculate discounts, add-to-cart items are marked through HTTP headers;

[0097] Amount-level isolation: exclude the amount of additional items when calculating the total order price.

[0098] In this embodiment, a data verification mechanism is established:

[0099] Deduplication hash lifecycle: Set a 15-minute expiration time (Redis TTL = 900 seconds) to prevent duplication across sessions

[0100] Discount isolation log tracking: records the isolated product ID and discount rule ID for subsequent conflict analysis.

[0101] In a possible embodiment, a test example of more than 5,000 vending machines is provided as follows:

[0102] The test environment is: AB testing of 5,000 vending machines for 30 consecutive days, with a data confidence level of p<0.01.

[0103] The test results are shown in Table 1:

[0104]

[0105]

[0106] Table 1 Comparison results before and after implementation

[0107] As shown in Table 1, the exposure rate of high-gross-profit items increased by 76.3%, the primary category balance compliance rate increased by 78.8%, the issue of discount conflicts was completely resolved, and the proportion of slow-moving items decreased by 52.7%. This demonstrates that the present invention effectively reduces the exposure of slow-moving items, increases the probability of recommending high-value items, avoids duplicate recommendations and discount conflicts, and improves the user experience.

[0108] In summary, the present invention constructs a mobile-side recommendation factor priority queue and a machine-side recommendation factor priority queue respectively based on the scenario differences between the mobile side and the machine side; based on the mobile-side recommendation factor priority queue, the recommendation factors are called in sequence to generate an initial candidate product pool; the initial candidate product pool is filtered and screened by setting multi-dimensional filtering rules to obtain a filtered candidate product pool; the filtered candidate product pool is checked for the number of first-level categories; if the number of first-level categories does not meet the minimum quantity requirement, the missing category products are re-screened and supplemented in the reverse order of the mobile-side recommendation factor priority queue to obtain a classified balanced product pool; the classified balanced product list is deduplicated and treated with preferential isolation to obtain a final product recommendation pool. The present invention reduces the exposure of slow-moving products, increases the probability of recommending high-value products, and improves the accuracy of recommendations through multi-factor priority and dynamic filling. The present invention ensures product diversity, avoids repeated recommendations, and optimizes user experience through forced balancing and deduplication mechanisms of category proportions. The present invention reduces losses caused by invalid recommendations and preferential conflicts and reduces operating costs through blacklist filtering and preferential isolation rules. The present invention solves the problems in the prior art such as a high proportion of slow-moving products, a single recommendation factor, uneven classification coverage, and poor user experience.

[0109] It should be noted that the method of the embodiments of the present disclosure can be performed by a single device, such as a computer or server. The method of the embodiments of the present disclosure can also be applied in a distributed scenario, where multiple devices cooperate to perform the method. In such a distributed scenario, one of the multiple devices may only perform one or more steps of the method of the embodiments of the present disclosure, and the multiple devices will interact with each other to complete the method.

[0110] It should be noted that the above description is limited to some embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in an order different from that described in the above embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0111] Example 2

[0112] See also Figure 5 Embodiment 2 of the present invention further provides a device for recommending additional items for vending machines based on multiple factors and dynamic classification, including:

[0113] Initial candidate product pool generation module 001 is used to construct a mobile terminal recommendation factor priority queue and a machine terminal recommendation factor priority queue based on the scenario differences between the mobile terminal and the machine terminal, respectively; based on the mobile terminal recommendation factor priority queue, recommendation factors are sequentially called to generate an initial candidate product pool;

[0114] The filtered candidate product pool acquisition module 002 is used to filter and screen the initial candidate product pool by setting multi-dimensional filtering rules to obtain a filtered candidate product pool;

[0115] The module 003 for obtaining a balanced product list by category is used to verify the number of first-level categories in the filtered candidate product pool; if the number of first-level categories does not meet the minimum quantity requirement, the module re-screens and supplements the missing category products in the reverse order of the mobile terminal recommendation factor priority queue to obtain a balanced product pool by category;

[0116] The final product recommendation pool acquisition module 004 is used to perform product deduplication and discount isolation processing on the classified balanced product list to obtain a final product recommendation pool.

[0117] In this embodiment, in the initial candidate product pool generation module 001, the recommendation factor priority queue of the mobile terminal is: manual configuration > user behavior > complementary products > product attributes > city hot sales;

[0118] The machine-side recommendation factor priority queue is manually configured as follows: > inventory turnover rate > complementary products > product attributes > city hot sales.

[0119] In this embodiment, in the initial candidate product pool generating module 001, in the process of sequentially calling recommendation factors based on the mobile terminal recommendation factor priority queue to generate the initial candidate product pool:

[0120] The candidate products for the recommendation factor are determined through a comprehensive scoring formula, and each non-terminal recommendation factor only recommends one product each time it is called. Unfilled positions are dynamically supplemented by the city's hot-selling factors.

[0121] The comprehensive scoring formula is:

[0122] S=α·M+β·U+γ·C+δ·A+ε·H

[0123] Where S is the total score; M is the manually configured weight, M=1 for configured products, otherwise M=0; U is the user behavior score; C is the correlation between complementary products; A is the product attribute score; H is the city's hot-selling coefficient; α, β, γ, δ, and ε are all weight coefficients, and they satisfy α>β>γ>δ>ε.

[0124] In this embodiment, in the filtered candidate product pool acquisition module 002, during the process of filtering the initial candidate product pool using the set multi-dimensional filtering rules:

[0125] The multi-dimensional filtering rules set include: goods to be paid, goods with an original price of ≤ 4 yuan, and goods in specific categories.

[0126] In this embodiment, the initial candidate product pool generation module 001 calculates a dynamic hot-selling pool based on basic sales volume and new product weighting strategy. The dynamic hot-selling pool provides real-time data support for the city hot-selling factor. The calculation formula of the dynamic hot-selling pool is:

[0127]

[0128] Where N is the monthly sales volume; d is the date; μ is the attenuation coefficient, usually 0.05; sale d is the sales volume on the past d day; e -μ·(30-d) is the exponential decay factor;

[0129]

[0130] Where N f is the final sales volume; θ is the weighting coefficient, which is 0.5; t is the number of days on the shelf; T=7 is the new product protection period.

[0131] It should be noted that the information interaction, execution process, etc. between the modules of the above-mentioned system are based on the same concept as the method embodiment in Example 1 of the present application, and the technical effects they bring are the same as those of the method embodiment of the present application. For specific contents, please refer to the description in the method embodiment shown above in the present application, and no further details will be given here.

[0132] Example 3

[0133] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, in which a program code for a method for recommending additional purchases of vending machine products based on multiple factors and dynamic classification is stored. The program code includes instructions for executing embodiment 1 or any possible implementation thereof.

[0134] Computer-readable storage media can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0135] Example 4

[0136] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0137] The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the vending machine product additional purchase recommendation method based on multiple factors and dynamic classification according to Example 1 or any possible implementation thereof.

[0138] Specifically, the processor can be implemented by hardware or by software. When implemented by hardware, the processor can be a logic circuit, an integrated circuit, etc.; when implemented by software, the processor can be a general-purpose processor, which is implemented by reading software code stored in a memory. The memory can be integrated into the processor or located outside the processor and exist independently.

[0139] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable systems. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website, computer, server or data center to another website, computer, server or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode.

[0140] Obviously, those skilled in the art will appreciate that the various modules or steps of the present invention described above can be implemented using a general-purpose computing system. They can be centralized on a single computing system or distributed across a network of multiple computing systems. Alternatively, they can be implemented using program code executable by a computing system, and thus, they can be stored in a storage system and executed by the computing system. In some cases, the steps shown or described herein can be performed in a different order than that shown, or they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0141] Although the present invention has been described in detail above using general descriptions and specific embodiments, it will be apparent to those skilled in the art that modifications and improvements may be made thereto. Therefore, such modifications and improvements, without departing from the spirit of the present invention, are intended to be within the scope of protection claimed herein.

Claims

1. A vending machine product add-on recommendation method based on multiple factors and dynamic classification, characterized by: include: Based on the differences in scenarios between mobile and machine terminals, we build a priority queue of recommendation factors for mobile terminals and a priority queue of recommendation factors for machine terminals respectively; Based on the mobile terminal recommendation factor priority queue, recommendation factors are called in sequence to generate an initial candidate product pool; Filtering the initial candidate product pool by setting multi-dimensional filtering rules to obtain a filtered candidate product pool; Performing a first-level classification quantity check on the filtered candidate product pool; If the number of first-level categories does not meet the minimum requirement, re-screen and supplement the missing category products in the reverse order of the mobile terminal recommendation factor priority queue to obtain a balanced category product pool; The classified balanced product list is subjected to product deduplication and discount isolation processing to obtain a final product recommendation pool.

2. The method for recommending additional products for vending machines based on multiple factors and dynamic classification according to claim 1, characterized in that: The priority queue of recommendation factors on the mobile terminal is: manual configuration > user behavior > complementary products > product attributes > city hot sales; The machine-side recommendation factor priority queue is manually configured as follows: > inventory turnover rate > complementary products > product attributes > city hot sales.

3. The method for recommending additional products for vending machines based on multiple factors and dynamic classification according to claim 2, characterized in that: In the process of sequentially calling recommendation factors based on the mobile terminal recommendation factor priority queue to generate the initial candidate product pool: The candidate products for the recommendation factor are determined through a comprehensive scoring formula, and each non-terminal recommendation factor only recommends one product each time it is called. Unfilled positions are dynamically supplemented by the city's hot-selling factors. The comprehensive scoring formula is: S=α·M+β·U+γ·C+δ·A+ε·H Where S is the total score; M is the manually configured weight, M=1 for configured products, otherwise M=0; U is the user behavior score; C is the correlation between complementary products; A is the product attribute score; H is the city's hot-selling coefficient; α, β, γ, δ, and ε are all weight coefficients, and they satisfy α>β>γ>δ>ε.

4. The method for recommending additional products for vending machines based on multiple factors and dynamic classification according to claim 3 is characterized in that: In the process of filtering the initial candidate product pool by setting the multi-dimensional filtering rules: The multi-dimensional filtering rules set include: goods to be paid, goods with an original price of ≤ 4 yuan, and goods in specific categories.

5. The method for recommending additional products for vending machines based on multiple factors and dynamic classification according to claim 4 is characterized in that: Based on the basic sales volume and the new product weighting strategy, a dynamic hot-selling pool is calculated; the dynamic hot-selling pool provides real-time data support for the city's hot-selling factors; the calculation formula of the dynamic hot-selling pool is: Where N is the monthly sales volume; d is the date; μ is the attenuation coefficient; sale d is the sales volume on the past d day; e -μ·(30-d) is the exponential decay factor; Where N f is the final sales volume; θ is the weighting coefficient; t is the number of days on the shelf; T is the new product protection period.

6. A device for recommending additional products for vending machines based on multiple factors and dynamic classification, characterized in that: include: An initial candidate product pool generation module is configured to construct a mobile-side recommendation factor priority queue and a machine-side recommendation factor priority queue based on the scenario differences between the mobile side and the machine side, respectively; and to sequentially call recommendation factors based on the mobile-side recommendation factor priority queue to generate an initial candidate product pool; A filtered candidate commodity pool acquisition module is used to filter and screen the initial candidate commodity pool by setting multi-dimensional filtering rules to obtain a filtered candidate commodity pool; A module for obtaining a classified balanced product list is used to verify the quantity of the first-level classification of the filtered candidate product pool; If the number of first-level categories does not meet the minimum requirement, re-screen and supplement the missing category products in the reverse order of the mobile terminal recommendation factor priority queue to obtain a balanced category product pool; The final product recommendation pool acquisition module is used to perform product deduplication and discount isolation processing on the classified balanced product list to obtain the final product recommendation pool.

7. The device for recommending additional products for vending machines based on multiple factors and dynamic classification according to claim 6, characterized in that: In the initial candidate product pool generation module, the mobile terminal's recommendation factor priority queue is: manual configuration > user behavior > complementary products > product attributes > city hot sales; The machine-side recommendation factor priority queue is manually configured as follows: > inventory turnover rate > complementary products > product attributes > city hot sales.

8. The device for recommending additional products for vending machines based on multiple factors and dynamic classification according to claim 7, characterized in that: In the initial candidate product pool generation module, in the process of sequentially calling recommendation factors based on the mobile terminal recommendation factor priority queue to generate the initial candidate product pool: The candidate products for the recommendation factor are determined through a comprehensive scoring formula, and each non-terminal recommendation factor only recommends one product each time it is called. Unfilled positions are dynamically supplemented by the city's hot-selling factors. The comprehensive scoring formula is: S=α·M+β·U+γ·C+δ·A+ε·H Where S is the total score; M is the manually configured weight, M=1 for configured products, otherwise M=0; U is the user behavior score; C is the correlation between complementary products; A is the product attribute score; H is the city's hot-selling coefficient; α, β, γ, δ, and ε are all weight coefficients, and they satisfy α>β>γ>δ>ε.

9. The device for recommending additional products for vending machines based on multiple factors and dynamic classification according to claim 8, characterized in that: In the filtered candidate product pool acquisition module, during the process of filtering the initial candidate product pool using the set multi-dimensional filtering rules: The multi-dimensional filtering rules set include: goods to be paid, goods with an original price of ≤ 4 yuan, and goods in specific categories.

10. The device for recommending additional products for vending machines based on multiple factors and dynamic classification according to claim 9, characterized in that: In the initial candidate product pool generation module, a dynamic hot-selling pool is calculated based on basic sales volume and new product weighting strategy; the dynamic hot-selling pool provides real-time data support for the city hot-selling factor; The calculation formula of the dynamic hot-selling pool is: Where N is the monthly sales volume; d is the date; μ is the attenuation coefficient; sale d is the sales volume on the past d day; e -μ·(30-d) is the exponential decay factor; Where N f is the final sales volume; θ is the weighting coefficient; t is the number of days on the shelf; T is the new product protection period.

Images