Retail cabinet replenishment analysis method and system based on multi-dimensional decision matrix

Abstract

The invention discloses a retail cabinet replenishment analysis method and system based on a multi-dimensional decision matrix, and the method comprises the steps: collecting the place characteristics of a retail cabinet, the user characteristics of a service object, and local urban hot commodity data, and constructing a multi-dimensional decision matrix; completing commodity compliance screening and recommendation based on the multi-dimensional decision matrix, and determining a candidate commodity set of a retail cabinet adapted to the current scene and the user demand; the shelf display positions of the candidate commodities are adjusted through a dynamic distribution rule in combination with the sales performance and attribute characteristics of the candidate commodities; according to the historical sales frequency of the commodities, commodity categories are divided by adopting a dynamic classification algorithm, and differential replenishment cycles are formulated for different categories of commodities; and predicting a commodity replenishment demand by using the sales prediction model, and planning an optimal delivery path in combination with a path optimization algorithm. The problems that due to static display of a traditional retail cabinet, the space utilization rate is low, stock-out or overstock is caused by rigid replenishment cycle, logistics distribution resources are wasted, and compliance management response is lagged are solved.

Description

Technical Field

[0001] This invention relates to the field of unmanned vending equipment technology, specifically to a retail cabinet replenishment analysis method and system based on a multi-dimensional decision matrix. Background Technology

[0002] With the booming development of the smart retail industry, unmanned retail kiosks, with their advantages of flexible deployment and 24-hour service, have been widely used in diverse scenarios such as transportation hubs, hospitals, and office parks, becoming an important supplementary form of retail. However, current retail kiosks still have many technical shortcomings in replenishment and operation management, making it difficult to adapt to dynamic market demands.

[0003] First, the rigid product display model leads to low space utilization. Traditional retail counters employ a static display strategy with fixed tiers, failing to optimize layout based on product characteristics and demand differences. This results in popular items being inaccessible while slow-moving goods occupy prime locations, severely impacting sales conversion. Second, the lack of flexibility in replenishment cycles leads to an imbalance between supply and demand. Existing solutions often use ABC classification methods with fixed cycles or requiring manual threshold adjustments, failing to respond to dynamic factors such as sales fluctuations and seasonal changes. This results in persistently high stockout rates for high-frequency items. Industry data shows that traditional retail suffers from high stockout rates due to untimely replenishment, while low-frequency items are prone to inventory buildup.

[0004] Furthermore, logistics and distribution resource allocation is inefficient. The dispersed deployment of retail kiosks means that the existing single-point independent delivery model lacks task aggregation and route optimization, resulting in high empty vehicle rates and significant resource waste. Additionally, the lack of a replenishment priority determination mechanism prolongs response time. Moreover, compliance management is lagging and risk control is insufficient. Different scenarios have clear restrictions on the sale of goods, but the current model relies on manual inspections to identify prohibited items, failing to keep pace with policy changes in real time and easily leading to compliance risks. Summary of the Invention

[0005] To address these issues, this invention provides a retail cabinet replenishment analysis method and system based on a multi-dimensional decision matrix, which solves the problems of low space utilization due to static display of traditional retail cabinets, rigid replenishment cycles leading to stockouts or backlogs, waste of logistics and distribution resources, and delayed compliance management response.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a retail counter replenishment analysis method based on a multi-dimensional decision matrix, comprising the following steps:

[0007] Collect data on the location characteristics of retail counters, the user characteristics of the service recipients, and the local city's best-selling products, and construct a multi-dimensional decision matrix that integrates the location characteristics, the user characteristics of the service recipients, and the local city's best-selling products data;

[0008] Based on the multi-dimensional decision matrix, the product compliance screening and recommendation are completed, and a set of candidate products for the retail cabinet that are suitable for the current scenario and user needs are determined.

[0009] Based on the sales performance and attribute characteristics of candidate products, the shelf display positions of candidate products are adjusted through dynamic allocation rules;

[0010] Based on the historical sales frequency of products, a dynamic classification algorithm is used to divide products into categories, and differentiated replenishment cycles are set for different product categories;

[0011] The sales forecasting model is used to predict the replenishment demand for goods, and the optimal delivery route is planned by combining the route optimization algorithm.

[0012] As a preferred approach for retail counter replenishment analysis based on a multi-dimensional decision matrix, the formula for constructing the multi-dimensional decision matrix is ​​as follows:

[0013]

[0014] In the formula, For product matching score matrix, For location feature adaptation matrix, For user feature matching matrix, A matrix adapted for top-selling cities. Let be the weight coefficients of the three matrices, and satisfy . .

[0015] As a preferred solution for retail cabinet replenishment analysis based on a multi-dimensional decision matrix, the location feature adaptation matrix... elements Represents the fit coefficient between the i-th type of goods and the j-th type of location, when This indicates that the goods are prohibited from sale in location type j. This indicates that the goods in this category are encouraged to be displayed in Category j locations.

[0016] As a preferred solution for retail counter replenishment analysis based on a multi-dimensional decision matrix, the dynamic update formula for the multi-dimensional decision matrix is:

[0017]

[0018] In the formula, For the updated multi-dimensional decision matrix, , , These are the updated location feature adaptation matrix, user feature matching matrix, and city best-selling product adaptation matrix; the update trigger conditions include changes in location feature information, iteration of user feature information, updates to city best-selling product data, and adjustments to policies on prohibited products.

[0019] As a preferred solution for retail shelf replenishment analysis based on a multi-dimensional decision matrix, the formula for implementing the dynamic allocation rule through a hierarchical weight calculation model during the process of adjusting the shelf display position of candidate products using dynamic allocation rules is as follows:

[0020]

[0021] In the formula, These are the weight coefficients of the specified influencing factors, and they satisfy the following conditions: ; These are the characteristic parameters related to product sales. These are characteristic parameters related to the profitability of the product. These are the time-sensitive characteristic parameters of the product; The value represents the actual sales quantity of the product within the preset statistical period. The value is the ratio of profit from product sales to sales revenue. The value is the number of remaining valid days for the product from the current date to the expiration date.

[0022] As a preferred solution for retail counter replenishment analysis based on a multi-dimensional decision matrix, the dynamic classification algorithm is the ABC dynamic classification algorithm. The formula for classifying goods using the dynamic classification algorithm is as follows:

[0023]

[0024] In the formula, This represents the percentage of product sales frequency. This refers to the total historical sales volume of a single product. This is the sum of the historical total sales volume of all products;

[0025] when At that time, the goods were classified as Category A high-frequency consumer goods; when At that time, the goods were classified as Category B mid-frequency consumer goods; when At that time, the goods were classified as Category C low-frequency consumer goods;

[0026] In the formula, and The dynamic classification threshold, and The threshold value is adjusted based on the historical sales trends of all products.

[0027] As a preferred scheme for retail counter replenishment analysis based on a multi-dimensional decision matrix, the calculation formulas for the differentiated replenishment cycle are as follows:

[0028] Replenishment cycle for Category A goods: sky;

[0029] Replenishment cycle for Category B goods: sky

[0030] Replenishment cycle for Category C goods:

[0031]

[0032] In the formula, The replenishment cycle for Category C products, This represents the current inventory quantity of the product. This represents the average daily sales volume of the product.

[0033] As a preferred solution for retail counter replenishment analysis based on a multi-dimensional decision matrix, the sales forecasting model is constructed using an LSTM neural network, and the formula for the sales forecasting model is:

[0034]

[0035] In the formula, This refers to the predicted sales volume of goods within a preset period. For the mapping function of the LSTM neural network, For historical sales data of the product, Seasonal variation factors As a factor affecting scene fluctuations, These are the relevant factors that influence product sales.

[0036] As a preferred solution for retail cabinet replenishment analysis based on a multi-dimensional decision matrix, the path optimization algorithm employs the ant colony algorithm, combined with replenishment task aggregation rules to achieve delivery route planning. The task aggregation determination formula is as follows:

[0037]

[0038] In the formula, The straight-line distance between the two retail counters. and These are the geographical coordinates of the two retail counters. Set the radius for task aggregation.

[0039] As a preferred scheme for retail cabinet replenishment analysis based on multi-dimensional decision matrices, the user characteristics of the service recipients are normalized to construct a user feature vector. The normalization formula is as follows:

[0040]

[0041] In the formula, Let i be the normalized value of the i-th user feature. For the original value of this feature for a single user, This is the minimum value of the feature. This is the maximum value of the feature;

[0042] User feature vector is , These are normalized values ​​for age characteristics. Normalized values ​​for occupational characteristics To normalize the purchase frequency, This represents the normalized value of consumer preferences.

[0043] As a preferred approach to retail counter replenishment analysis based on a multi-dimensional decision matrix, the update formula for the city's best-selling product data is:

[0044]

[0045] In the formula, This is an updated collection of best-selling products. This is a collection of best-selling products before the update. This is a collection of the latest top 10 best-selling products. These are the weighting coefficients.

[0046] This invention also provides a retail counter replenishment analysis method based on a multi-dimensional decision matrix. The method includes:

[0047] The multi-dimensional decision matrix construction module is used to collect the location characteristics of retail counters, the user characteristics of service objects, and the local city's best-selling product data, and to construct a multi-dimensional decision matrix that integrates the location characteristics, the user characteristics of service objects, and the local city's best-selling product data;

[0048] The candidate product set determination module is used to complete the compliance screening and recommendation of products based on the multi-dimensional decision matrix, and determine the candidate product set of the retail cabinet that is suitable for the current scenario and user needs;

[0049] The candidate product display adjustment module is used to adjust the shelf display position of candidate products by combining their sales performance and attribute characteristics through dynamic allocation rules.

[0050] The replenishment cycle analysis module is used to classify products into categories based on their historical sales frequency using a dynamic classification algorithm, and to formulate differentiated replenishment cycles for different product categories.

[0051] The replenishment demand forecasting module is used to predict the replenishment demand of goods using a sales forecasting model and to plan the optimal delivery route using a route optimization algorithm.

[0052] The present invention has the following advantages:

[0053] First, this invention achieves the matching of products with scenarios and user needs through a multi-dimensional decision matrix, automatically filters suitable products and blocks prohibited categories to meet the compliance requirements of different scenarios; it dynamically adjusts the display position based on the core attributes of the products, allowing highly suitable products to occupy advantageous display resources and solving the problem of space waste caused by static displays.

[0054] Secondly, this invention avoids stockouts of high-frequency goods and inventory backlogs of low-frequency goods by combining adaptive replenishment cycles with sales forecasting, thereby reducing product spoilage costs. Through path planning algorithms and task aggregation strategies, it rationally integrates delivery resources, shortens replenishment response time, and reduces waste of logistics resources.

[0055] Third, this invention can flexibly adjust product selection, display, and replenishment strategies according to different venue attributes, user characteristics, seasons, and holidays, making it widely applicable. Relying on a screening mechanism and a real-time updated prohibited goods database, it can quickly respond to policy changes without manual inspections, thus avoiding the risk of violations.

[0056] Fourth, this invention automates and automates the entire process of product selection, display, replenishment, and delivery, reducing manual intervention and improving the overall operational efficiency of retail counters. Attached Figure Description

[0057] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0058] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0059] Figure 1 This is a schematic diagram of the retail counter replenishment analysis method based on a multi-dimensional decision matrix provided in an embodiment of the present invention;

[0060] Figure 2 This is a schematic diagram of the technical route of the retail cabinet replenishment analysis method based on a multi-dimensional decision matrix provided in the embodiments of the present invention;

[0061] Figure 3 This is a schematic diagram of the retail counter replenishment analysis system architecture based on a multi-dimensional decision matrix provided in an embodiment of the present invention. Detailed Implementation

[0062] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Example 1

[0064] See Figure 1 and Figure 2 Embodiment 1 of the present invention provides a retail counter replenishment analysis method based on a multi-dimensional decision matrix, comprising the following steps:

[0065] S1. Collect data on the location characteristics of the retail counters, the user characteristics of the service recipients, and the best-selling products in the local city. Construct a multi-dimensional decision matrix that integrates these three data points. Data collection is the foundation for building the multi-dimensional decision matrix. Location characteristics need to capture the attributes of the retail counter's location, such as hospitals, transportation hubs, and office parks, providing a basis for compliant screening and scenario adaptation. User characteristics focus on key information such as the age, occupation, purchase frequency, and consumption preferences of the service recipients. Data on best-selling products in the local city reflects the dynamic demand of the regional market, ensuring that product selection aligns with mass consumption trends. The multi-dimensional decision matrix breaks through the limitations of single-dimensional decision-making by integrating these three types of data, using matrix operations to achieve multi-dimensional matching of products with scenarios, users, and the market.

[0066] S2. Based on the multi-dimensional decision matrix, complete the compliance screening and recommendation of goods to determine the candidate product set for the retail cabinet that is suitable for the current scenario and user needs. The multi-dimensional decision matrix contains the adaptation rules and weight allocation of data in each dimension. In the compliance screening stage, the elements of the location feature adaptation matrix in the matrix are used to automatically filter out goods that do not meet the requirements of the scenario policy, ensuring operational compliance. In the product recommendation stage, the adaptation score of the user feature matching matrix and the market popularity data of the city best-selling adaptation matrix are combined to comprehensively score and rank all goods, and select the goods with high scores that meet the scenario compliance requirements, user needs and market popularity to form a candidate product set, avoiding the subjectivity and blindness of manual product selection.

[0067] S3. Combining the sales performance and attribute characteristics of candidate products, adjust their shelf display positions using dynamic allocation rules. Shelf display position directly affects product reach and sales conversion. Dynamic allocation rules evaluate product priority using quantitative indicators. Sales performance reflects the market acceptance of products, while attribute characteristics relate to operational revenue and loss costs. By constructing a hierarchical weight calculation model, these indicators are converted into calculable weight values. Shelf levels are allocated according to weight values. Products with higher weight values ​​are better able to meet user needs, generate higher revenue, and have lower loss risks; therefore, they are allocated to the prime display level, achieving optimal allocation of display resources and improving overall space efficiency.

[0068] S4. Based on historical sales frequency, a dynamic classification algorithm is used to categorize products, and differentiated replenishment cycles are established for different product categories. Historical sales frequency directly reflects the intensity of market demand. The dynamic classification algorithm calculates the proportion of total sales for each product, classifying them into three categories: high-frequency (Category A), medium-frequency (Category B), and low-frequency (Category C). The design logic of the differentiated replenishment cycle is to replenish on demand. Category A products have strong demand, requiring shorter replenishment cycles to avoid stockouts; Category B products have stable demand, using fixed-cycle replenishment to balance inventory and costs; Category C products have weak demand, with replenishment cycles determined based on the ratio of current inventory to average daily sales to avoid inventory buildup. Simultaneously, the classification threshold is dynamically adjusted according to changes in historical sales volume to ensure that the classification results always adapt to fluctuations in market demand.

[0069] S5. Utilize a sales forecasting model to predict product replenishment demand, and combine it with a path optimization algorithm to plan the optimal delivery route. The sales forecasting model leverages the time-series data processing advantages of LSTM neural networks. By learning the changing patterns of multiple influencing factors such as historical sales data, seasonal variations, and scenario fluctuations, it accurately predicts product sales within a preset period, providing data support for determining replenishment quantities and avoiding insufficient or excessive replenishment. The path optimization algorithm employs the swarm intelligence optimization characteristics of ant colony optimization, simulating the path selection logic of ants foraging. Combined with replenishment task aggregation rules, it integrates replenishment tasks within a certain radius, planning the delivery route with the shortest total mileage and least time consumption, reducing waste of delivery resources and improving replenishment response efficiency.

[0070] In this embodiment, in step S1, the formula for constructing the multi-dimensional decision matrix is:

[0071]

[0072] In the formula, For product matching score matrix, For location feature adaptation matrix, For user feature matching matrix, A matrix adapted for top-selling cities. Let be the weight coefficients of the three matrices, and satisfy . Weighting coefficients , , The settings reflect the priority of different dimensions' impact on product matching accuracy and can be adjusted according to the operational goals of the actual application scenario. For example, in scenarios with strict compliance requirements, the setting can be increased. Weighting. Three matching matrices score products from three dimensions: scenario compliance, user needs, and market popularity. The weighted sum of these matrices yields the product matching score matrix. This can intuitively reflect the comprehensive adaptability of each product, providing a quantitative basis for product selection decisions.

[0073] In this embodiment, in step S1, the location feature adaptation matrix elements Represents the fit coefficient between the i-th type of goods and the j-th type of location, when This indicates that the goods are prohibited from sale in location type j. This indicates that the goods in this category are encouraged to be displayed in Category j locations.

[0074] Specifically, the location feature adaptation matrix It is a tool for achieving scenario compliance and adaptation, and its elements The matching coefficient is determined through predefined scenario-product adaptation rules. The range of the adaptation coefficient [0,1] quantifies the degree of adaptation between the product and the location. A value of 0 clearly defines the prohibited sales boundary to avoid compliance risks; a high adaptation coefficient of ≥0.8 filters out products with core needs in the scenario, providing a clear standard for encouraging display and ensuring that the product and the scenario's functional positioning are highly consistent.

[0075] In one possible embodiment, in step S1, the dynamic update formula for the multi-dimensional decision matrix is:

[0076]

[0077] In the formula, For the updated multi-dimensional decision matrix, , , These are the updated location feature adaptation matrix, user feature matching matrix, and city best-selling product adaptation matrix; the update trigger conditions include changes in location feature information, iteration of user feature information, updates to city best-selling product data, and adjustments to policies on prohibited products.

[0078] Specifically, a dynamic update mechanism for the multi-dimensional decision matrix is ​​crucial for adapting to market and environmental changes. When venue attributes change, user group structure changes, best-selling products are iterated, or sales ban policies are updated, the corresponding adaptation matrix needs to be updated synchronously, such as... Adjust the scene-product adaptation coefficient. Replace the existing set of best-selling products. By keeping the weight coefficients constant and updating the input matrix, the product matching score matrix is ​​ensured. It always reflects the latest scenarios, user and market conditions, avoiding product selection bias caused by a rigid decision matrix.

[0079] In this embodiment, during step S3, when adjusting the shelf display position of candidate products through dynamic allocation rules, the formula for implementing the dynamic allocation rules through the hierarchical weight calculation model is as follows:

[0080]

[0081] In the formula, These are the weight coefficients of the specified influencing factors, and they satisfy the following conditions: ; These are the characteristic parameters related to product sales. These are characteristic parameters related to the profitability of the product. These are the time-sensitive characteristic parameters of the product; The value represents the actual sales quantity of the product within the preset statistical period. The value is the ratio of profit from product sales to sales revenue. The value is the number of remaining valid days for the product from the current date to the expiration date.

[0082] Specifically, weighting coefficients , , The settings balance the three core operational objectives of sales performance, profitability, and timeliness risk. Reflecting the market demand for goods, Reflecting the profit contribution of the product, The risk of product spoilage, along with the weighted sum of these three factors, yields a hierarchical weight P, which quantifies the priority of products occupying prime display resources. By calculating the P value in real time and adjusting display positions accordingly, it ensures that prime display resources are consistently allocated to products with high demand, high returns, and low spoilage, maximizing the commercial value of shelf space.

[0083] In this embodiment, in step S4, the dynamic classification algorithm is the ABC dynamic classification algorithm, and the formula for dividing product categories using the dynamic classification algorithm is:

[0084]

[0085] In the formula, This represents the percentage of product sales frequency. This refers to the total historical sales volume of a single product. This is the sum of the historical total sales volume of all products;

[0086] when At that time, the goods were classified as Category A high-frequency consumer goods; when At that time, the goods were classified as Category B mid-frequency consumer goods; when At that time, the goods were classified as Category C low-frequency consumer goods;

[0087] In the formula, and The dynamic classification threshold, and The threshold value is adjusted based on the historical sales trends of all products.

[0088] Specifically, the ABC dynamic classification algorithm is based on the principle that a few products contribute to the majority of sales, and classifies them by sales frequency ratio. Quantify the intensity of demand for a product. Historical total sales volume of a single product. Sum of all historical sales of all products ratio This directly reflects the importance of the product in the overall sales structure. Dynamic classification threshold. , It is not a fixed value, but rather an adaptive adjustment based on historical sales trends, such as overall sales growth or a surge in demand for a certain type of product. This ensures that the classification of products into categories A, B, and C always aligns with the actual sales structure, providing a basis for setting differentiated replenishment cycles.

[0089] In this embodiment, in step S4, the calculation formulas for the differentiated replenishment cycle are as follows:

[0090] Replenishment cycle for Category A goods: sky;

[0091] Replenishment cycle for Category B goods: sky

[0092] Replenishment cycle for Category C goods:

[0093]

[0094] In the formula, The replenishment cycle for Category C products, This represents the current inventory quantity of the product. This represents the average daily sales volume of the product.

[0095] Specifically, the design of differentiated replenishment cycles is based on the differences in the frequency of product demand, achieving a balance between inventory costs and stockout risk. Category A products are consumed frequently with consistently strong demand; a short replenishment cycle of 1 day can minimize stockouts. Category B products have stable demand; a 7-day replenishment cycle can meet daily needs while reducing delivery frequency and lowering costs. Category C products have weak demand; a fixed replenishment cycle could easily lead to inventory buildup, therefore, replenishment cycles are determined based on current inventory levels. Daily sales volume The ratio is used to calculate the replenishment cycle. This ensures that replenishment timing matches the rate of inventory depletion, enabling replenishment before inventory runs out and reducing the risk of overstocking and losses.

[0096] In this embodiment, in step S5, the sales forecasting model is constructed using an LSTM neural network, and the formula for the sales forecasting model is:

[0097]

[0098] In the formula, This refers to the predicted sales volume of goods within a preset period. For the mapping function of the LSTM neural network, For historical sales data of the product, Seasonal variation factors As a factor affecting scene fluctuations, These are the relevant factors that influence product sales.

[0099] Specifically, LSTM neural networks have the advantage of processing time-series data and capturing long-term dependencies, making them suitable for product sales prediction scenarios. Mapping function The training process learns the non-linear relationship between various influencing factors and sales volume. Historical sales data provides the basic patterns of sales changes. (Seasonal Changes) Capturing the cyclical influences of temperature, holidays, etc. (Scene fluctuations) Adapt to changes in passenger flow in different scenarios (such as increased passenger flow at transportation hubs during holidays). (Other factors) can include additional influencing factors such as promotional activities and competitor activities. Through multi-factor input, the model can output sales forecasts for a preset period. This provides a scientific basis for calculating replenishment quantities and avoids blind replenishment.

[0100] In this embodiment, in step S5, the path optimization algorithm adopts the ant colony algorithm, combined with the replenishment task aggregation rule to realize delivery route planning. The task aggregation determination formula is:

[0101]

[0102] In the formula, The straight-line distance between the two retail counters. and These are the geographical coordinates of the two retail counters. Set the radius for task aggregation.

[0103] Specifically, the ant colony algorithm simulates the behavior of ants releasing pheromones and choosing paths based on those pheromones during foraging to achieve a globally optimal path search. In replenishment path planning, it first uses a task aggregation decision formula to filter out paths that are closer (≤... The system aggregates the replenishment tasks of retail kiosks into delivery units, reducing the number of round trips for delivery vehicles. Then, using the geographical coordinates of the delivery units as nodes and delivery mileage and time as objective functions, the ant colony algorithm guides the search for the optimal path through a pheromone update mechanism. Ultimately, it plans the delivery route that covers all aggregated tasks and has the lowest total cost, improving delivery efficiency and reducing logistics costs.

[0104] In one possible embodiment, in step S1, the user features of the service object are normalized to construct a user feature vector. The normalization formula is:

[0105]

[0106] In the formula, Let i be the normalized value of the i-th user feature. For the original value of this feature for a single user, This is the minimum value of the feature. This is the maximum value of the feature;

[0107] User feature vector is , These are normalized values ​​for age characteristics. Normalized values ​​for occupational characteristics To normalize the purchase frequency, This represents the normalized value of consumer preferences.

[0108] Specifically, raw user characteristic data, such as age and purchase frequency, have different units of measurement, and directly using them in calculations can lead to weight imbalances. Normalization processes address this by adjusting the raw values... Mapping to the [0,1] interval eliminates the influence of dimensions, placing different features on the same order of magnitude and ensuring that each feature plays a balanced role in subsequent matrix operations. The constructed user feature vector V integrates four core dimensions: age, occupation, purchase frequency, and consumption preferences, forming a comprehensive quantitative description of user needs. This provides standardized data for the construction of the user feature matching matrix B, improving the matching accuracy between products and user needs.

[0109] In one possible embodiment, in step S1, the update formula for the city's best-selling product data is:

[0110]

[0111] In the formula, This is an updated collection of best-selling products. This is a collection of best-selling products before the update. This is a collection of the latest top 10 best-selling products. These are the weighting coefficients.

[0112] Specifically, the updating of city best-selling product data needs to balance stability and timeliness to avoid fluctuations in product selection due to frequent updates. The weighting coefficient α (ranging from 0 to 1) controls the proportion of historical best-selling data retained, while (1-α) controls the proportion of the latest best-selling data introduced. A larger α results in greater stability of the best-selling product set; a smaller α results in greater sensitivity to new market trends. The updated set is then weighted and merged. With the latest capture It retains popular products that have been proven in the market, while also incorporating newly emerging best-selling categories in a timely manner, ensuring that the city best-selling matching matrix always keeps up with market dynamics, while avoiding drastic fluctuations in product selection that could affect the user experience.

[0113] The application scenarios of this invention are as follows:

[0114] Healthcare scenarios:

[0115] Covering locations such as general hospitals, specialized hospitals, community health service centers, and nursing homes, retail counters can be deployed in areas such as outpatient halls, inpatient floors, emergency room corridors, and rehabilitation areas. In this scenario, the location feature matching matrix automatically identifies medical and health attributes, filtering out non-compliant categories such as high-sugar beverages, alcoholic products, and spicy snacks. Simultaneously, it encourages the display of sugar-free beverages, high-protein nutritional foods, portable daily necessities (such as disposable masks and wet wipes), and chronic disease management-related auxiliary products with a matching coefficient ≥0.8. Based on user profiles (primarily elderly people, patients, and medical staff), the user feature matching matrix prioritizes recommending low-sugar, low-fat, and easily digestible foods and small-sized daily necessities to meet the nutritional needs of patients during recovery and the immediate replenishment needs of caregivers. Given the urgency of the emergency room, the path optimization algorithm prioritizes replenishment tasks to ensure uninterrupted supply of emergency goods (such as functional beverages and energy bars). Simultaneously, the LSTM sales forecasting model dynamically adjusts replenishment quantities based on seasonal variations (such as increased demand for electrolyte water during summer heat) and flu season fluctuations to avoid stockouts or inventory buildup. Furthermore, when compliance policies in the healthcare sector are updated (such as new restrictions on the sale of certain food items), the multi-dimensional decision matrix is ​​updated in real time, simultaneously adjusting product selection rules to ensure operational compliance.

[0116] Transportation hub scenario:

[0117] Suitable for high-traffic, diverse locations such as high-speed rail stations, airports, subway stations, long-distance bus stations, and ports, the retail kiosks can be deployed in core areas such as waiting halls, airport terminals, transfer passages, and exits. The core requirement of this scenario is to adapt to the high-volume, high-frequency immediate consumption and travel replenishment needs of passengers, while also meeting the diverse preferences of travelers from different regions. A location-specific adaptation matrix filters out prohibited items such as flammable and explosive materials and prohibited knives. Combined with a city-wide best-selling data module, it captures local specialty products (such as local snacks and cultural and creative products) and nationally popular travel essentials (such as disposable chargers, portable rain gear, and noise-canceling earplugs) in real time and loads them into prime display areas. During peak travel periods such as holidays, Spring Festival travel rush, and summer travel rush, the ABC dynamic classification algorithm automatically upgrades high-frequency consumer goods such as beverages, snacks, and convenience foods to category A, shortening the replenishment cycle to one day and ensuring sufficient supply. During off-peak periods, the classification threshold is dynamically adjusted to avoid inventory waste. The route optimization algorithm aggregates replenishment tasks from multiple retail counters within a 3km radius of the same transportation hub, plans the optimal delivery route, reduces the frequency of delivery vehicle round trips, and improves replenishment response speed in high-traffic scenarios. Meanwhile, data on best-selling urban products is continuously iterated through a weighted update formula, retaining classic best-selling categories while promptly incorporating new essential travel items to adapt to changing traveler consumption trends.

[0118] Office and park scenarios:

[0119] This includes locations with high concentrations of working professionals, such as urban office buildings, industrial parks, science parks, and startup incubators. Retail kiosks can be deployed on office floors, around park cafeterias, and in shared spaces. Users in this scenario are primarily white-collar workers and corporate employees, whose consumption needs are concentrated on convenient breakfasts, snacks during work breaks, and healthy afternoon tea. They prefer healthy drinks, low-sugar snacks, convenient lunch alternatives (such as instant rice and sandwiches), and energy drinks (such as black coffee and energy drinks). A multi-dimensional decision matrix adapts to the "office commuting" attribute based on location characteristics, prioritizing the display of healthy and convenient products with high suitability. Combining user profiles (age concentrated between 25-45, valuing efficiency and health, and purchasing frequency concentrated during morning rush hour and lunch break), and through user feature vector analysis (occupational characteristics: working professionals, high purchase frequency, and a preference for healthy, low-fat products), it recommends products that meet their needs. To address the sales differences between weekdays and weekends, the dynamic replenishment cycle will be flexibly adjusted: on weekdays, coffee, yogurt, and meal replacement products will be designated as Category A items and replenished daily; on weekends, due to reduced customer traffic, the replenishment frequency of some items will be automatically reduced to avoid inventory backlog. The dynamic shelf allocation rules will place high-selling, high-profit-margin energy drinks and portable snacks on prime shelves (such as shelves 1-2) for easy access by working professionals; simultaneously, based on product shelf-life parameters, short-shelf-life fresh food items will be prioritized for display to reduce the risk of spoilage.

[0120] Campus scene:

[0121] Covering primary and secondary schools, vocational schools, and universities, retail kiosks can be deployed in areas such as teaching building corridors, dormitory buildings, playgrounds, and library lobbies. This scenario must strictly adhere to campus food hygiene and safety policies while adapting to the consumption characteristics and diverse needs of student groups. The location feature adaptation matrix automatically filters out tobacco and alcohol, high-sugar and high-fat "junk food," and products unsuitable for students' age groups, encouraging the display of stationery, low-sugar drinks, healthy snacks (such as nuts and dried fruit), and learning aids (such as sticky notes and pens). Addressing the different needs of students at different educational levels, primary and secondary school students focus on low-priced, small-sized snacks and stationery, with user feature vectors emphasizing age (12-18 years old) and consumption preferences (preference for fun packaging and affordable goods); university students are encouraged to purchase meal replacement foods, functional beverages, and supplements for postgraduate entrance exam preparation, adapting to the learning and lifestyle pace of university students. The ABC dynamic classification algorithm categorizes high-frequency consumer goods such as stationery, bottled water, and common snacks into Category A, requiring daily replenishment; seasonal goods (such as winter beverages) are categorized into Category B, requiring weekly replenishment; and niche stationery and special categories are categorized into Category C, with replenishment cycles determined by the ratio of current inventory to average daily sales. The LSTM sales forecasting model incorporates fluctuation factors related to school seasons, exam seasons, and winter / summer breaks to predict replenishment demand. For example, during exam seasons, replenishment of energy drinks and foods is increased, while during winter / summer breaks, replenishment frequency is reduced to avoid inventory buildup. When campus sales bans are updated or student consumption preferences change (such as the popularity of certain healthy snacks), the multi-dimensional decision matrix is ​​updated in real time, adjusting product selection and display rules accordingly.

[0122] It should be noted that the method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method in this embodiment, and the multiple devices will interact with each other to complete the retail cabinet replenishment analysis method based on a multi-dimensional decision matrix.

[0123] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0124] Example 2

[0125] See Figure 3Embodiment 2 of the present invention also provides a retail counter replenishment analysis method based on a multi-dimensional decision matrix. The retail counter replenishment analysis method based on a multi-dimensional decision matrix described in the above embodiments includes:

[0126] The multi-dimensional decision matrix construction module 100 is used to collect the location characteristics of the retail cabinet, the user characteristics of the service objects, and the local city's best-selling product data, and to construct a multi-dimensional decision matrix that integrates the location characteristics, the user characteristics of the service objects, and the local city's best-selling product data.

[0127] The candidate product set determination module 200 is used to complete the compliance screening and recommendation of products based on the multi-dimensional decision matrix, and determine the candidate product set of the retail cabinet that is suitable for the current scenario and user needs;

[0128] The candidate product display adjustment module 300 is used to adjust the shelf display position of candidate products by combining their sales performance and attribute characteristics through dynamic allocation rules.

[0129] The replenishment cycle analysis module 400 is used to classify products into categories based on their historical sales frequency using a dynamic classification algorithm, and to formulate differentiated replenishment cycles for different product categories.

[0130] The replenishment demand forecasting module 500 is used to predict the replenishment demand of goods using a sales forecasting model and to plan the optimal delivery route using a route optimization algorithm.

[0131] In this embodiment, the construction formula of the multi-dimensional decision matrix in the multi-dimensional decision matrix construction module 100 is as follows:

[0132]

[0133] In the formula, For product matching score matrix, For location feature adaptation matrix, For user feature matching matrix, A matrix adapted for top-selling cities. Let be the weight coefficients of the three matrices, and satisfy . .

[0134] In this embodiment, the location feature adaptation matrix in the multi-dimensional decision matrix construction module 100 elements Represents the fit coefficient between the i-th type of goods and the j-th type of location, when This indicates that the goods are prohibited from sale in location type j. This indicates that the goods in this category are encouraged to be displayed in Category j locations.

[0135] In this embodiment, the dynamic update formula of the multi-dimensional decision matrix in the multi-dimensional decision matrix construction module 100 is as follows:

[0136]

[0137] In the formula, For the updated multi-dimensional decision matrix, , , These are the updated location feature adaptation matrix, user feature matching matrix, and city best-selling product adaptation matrix; the update trigger conditions include changes in location feature information, iteration of user feature information, updates to city best-selling product data, and adjustments to policies on prohibited products.

[0138] In this embodiment, the formula for implementing the dynamic allocation rule in the candidate product display adjustment module 300 through the hierarchical weight calculation model is as follows:

[0139]

[0140] In the formula, These are the weight coefficients of the specified influencing factors, and they satisfy the following conditions: ; These are the characteristic parameters related to product sales. These are characteristic parameters related to the profitability of the product. These are the time-sensitive characteristic parameters of the product; The value represents the actual sales quantity of the product within the preset statistical period. The value is the ratio of profit from product sales to sales revenue. The value is the number of remaining valid days for the product from the current date to the expiration date.

[0141] In this embodiment, the dynamic classification algorithm in the replenishment cycle analysis module 400 is the ABC dynamic classification algorithm. The formula for classifying product categories using the dynamic classification algorithm is as follows:

[0142]

[0143] In the formula, This represents the percentage of product sales frequency. This refers to the total historical sales volume of a single product. This is the sum of the historical total sales volume of all products;

[0144] when At that time, the goods were classified as Category A high-frequency consumer goods; when At that time, the goods were classified as Category B mid-frequency consumer goods; when At that time, the goods were classified as Category C low-frequency consumer goods;

[0145] In the formula, and The dynamic classification threshold, and The threshold value is adjusted based on the historical sales trends of all products.

[0146] In this embodiment, the calculation formulas for the differentiated replenishment cycle in the replenishment cycle analysis module 400 are as follows:

[0147] Replenishment cycle for Category A goods: sky;

[0148] Replenishment cycle for Category B goods: sky

[0149] Replenishment cycle for Category C goods:

[0150]

[0151] In the formula, The replenishment cycle for Category C products, This represents the current inventory quantity of the product. This represents the average daily sales volume of the product.

[0152] In this embodiment, in the replenishment demand prediction module 500, the sales prediction model is constructed using an LSTM neural network, and the formula for the sales prediction model is:

[0153]

[0154] In the formula, This refers to the predicted sales volume of goods within a preset period. For the mapping function of the LSTM neural network, For historical sales data of the product, Seasonal variation factors As a factor affecting scene fluctuations, These are the relevant factors that influence product sales.

[0155] In this embodiment, the replenishment demand prediction module 500 uses an ant colony algorithm for path optimization, combined with replenishment task aggregation rules to achieve delivery route planning. The task aggregation determination formula is as follows:

[0156]

[0157] In the formula, The straight-line distance between the two retail counters. and These are the geographical coordinates of the two retail counters. Set the radius for task aggregation.

[0158] In one possible embodiment, in the multi-dimensional decision matrix construction module 100, the user features of the service object are normalized to construct a user feature vector, and the normalization formula is:

[0159]

[0160] In the formula, Let i be the normalized value of the i-th user feature. For the original value of this feature for a single user, This is the minimum value of the feature. This is the maximum value of the feature;

[0161] User feature vector is , These are normalized values ​​for age characteristics. Normalized values ​​for occupational characteristics To normalize the purchase frequency, This represents the normalized value of consumer preferences.

[0162] In one possible embodiment, the update formula for the city's best-selling product data in the multi-dimensional decision matrix construction module 100 is:

[0163]

[0164] In the formula, This is an updated collection of best-selling products. This is a collection of best-selling products before the update. This is a collection of the latest top 10 best-selling products. These are the weighting coefficients.

[0165] It should be noted that the information interaction and execution process between the modules of the above system are based on the same concept as the method embodiment in Embodiment 1 of this application, and the resulting technical effects are the same as those in the method embodiment of this application. For details, please refer to the description in the method embodiment shown above in this application, and it will not be repeated here.

[0166] Example 3

[0167] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium storing program code for a retail counter replenishment analysis method based on a multi-dimensional decision matrix. The program code includes instructions for executing the retail counter replenishment analysis method based on a multi-dimensional decision matrix as described in Embodiment 1 or any possible implementation thereof.

[0168] Computer-readable storage media can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium 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)).

[0169] Example 4

[0170] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0171] 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 can execute the retail counter replenishment analysis method based on a multi-dimensional decision matrix according to Embodiment 1 or any possible implementation thereof by calling the program instructions.

[0172] Specifically, a processor can be implemented in hardware or software. When implemented in hardware, the processor can be a logic circuit, an integrated circuit, etc. When implemented in software, the processor can be a general-purpose processor that reads software code stored in memory. This memory can be integrated into the processor or located outside the processor and exist independently.

[0173] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as 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, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable system. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0174] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing systems. They can be centralized on a single computing system or distributed across a network of multiple computing systems. Optionally, they can be implemented using program code executable by a computing system, thereby storing them in a storage system for execution by the computing system. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0175] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A retail counter replenishment analysis method based on a multi-dimensional decision matrix, characterized in that, Includes the following steps: Collect data on the location characteristics of retail counters, the user characteristics of the service recipients, and the local city's best-selling products, and construct a multi-dimensional decision matrix that integrates the location characteristics, the user characteristics of the service recipients, and the local city's best-selling products data; Based on the multi-dimensional decision matrix, the product compliance screening and recommendation are completed, and a set of candidate products for the retail cabinet that are suitable for the current scenario and user needs are determined. Based on the sales performance and attribute characteristics of candidate products, the shelf display positions of candidate products are adjusted through dynamic allocation rules; Based on the historical sales frequency of products, a dynamic classification algorithm is used to divide products into categories, and differentiated replenishment cycles are set for different product categories; The sales forecasting model is used to predict the replenishment demand for goods, and the optimal delivery route is planned by combining the route optimization algorithm.

2. The retail counter replenishment analysis method based on a multi-dimensional decision matrix according to claim 1, characterized in that, The formula for constructing the multi-dimensional decision matrix is ​​as follows: In the formula, For product matching score matrix, For location feature adaptation matrix, For user feature matching matrix, A matrix adapted for best-selling cities. Let be the weight coefficients of the three matrices, and satisfy . ; The location feature adaptation matrix elements Represents the fit coefficient between the i-th type of goods and the j-th type of location, when This indicates that the goods are prohibited from sale in location type j. This indicates that the goods in question are encouraged for display in category j locations; The dynamic update formula for the multi-dimensional decision matrix is: In the formula, For the updated multi-dimensional decision matrix, , , These are the updated location feature adaptation matrix, user feature matching matrix, and city best-selling product adaptation matrix; the update trigger conditions include changes in location feature information, iteration of user feature information, updates to city best-selling product data, and adjustments to policies on prohibited products.

3. The retail counter replenishment analysis method based on a multi-dimensional decision matrix according to claim 1, characterized in that, In the process of adjusting the shelf display position of candidate products through dynamic allocation rules, the formula for implementing the dynamic allocation rules through the hierarchical weight calculation model is as follows: In the formula, These are the weight coefficients of the specified influencing factors, and they satisfy the following conditions: ; These are the characteristic parameters related to product sales. These are characteristic parameters related to the profitability of the product. These are the time-sensitive characteristic parameters of the product; The value represents the actual sales quantity of the product within a preset statistical period. The value is the ratio of profit from product sales to sales revenue. The value is the number of remaining valid days for the product from the current date to the expiration date.

4. The retail counter replenishment analysis method based on a multi-dimensional decision matrix according to claim 1, characterized in that, The dynamic classification algorithm is the ABC dynamic classification algorithm, and the formula for dividing product categories using the dynamic classification algorithm is as follows: In the formula, This represents the percentage of product sales frequency. This refers to the total historical sales volume of a single product. This is the sum of the historical total sales volume of all products; when At that time, the goods were classified as Category A high-frequency consumer goods; when At that time, the goods were classified as Category B mid-frequency consumer goods; when At that time, the goods were classified as Category C low-frequency consumer goods; In the formula, and The dynamic classification threshold, and The threshold value is adjusted based on the historical sales trends of all products.

5. The retail counter replenishment analysis method based on a multi-dimensional decision matrix according to claim 4, characterized in that, The formulas for calculating the differentiated replenishment cycle are as follows: Replenishment cycle for Category A goods: sky; Replenishment cycle for Category B goods: sky Restocking cycle for Category C goods: In the formula, The replenishment cycle for Category C products, This represents the current inventory quantity of the product. This represents the average daily sales volume of the product.

6. The retail counter replenishment analysis method based on a multi-dimensional decision matrix according to claim 1, characterized in that, The sales forecasting model is constructed using an LSTM neural network, and the formula for the sales forecasting model is as follows: In the formula, This refers to the predicted sales volume of goods within a preset period. For the mapping function of the LSTM neural network, For historical sales data of the product, Seasonal variation influencing factors As a factor affecting scene fluctuations, These are the relevant factors that influence product sales.

7. The retail counter replenishment analysis method based on a multi-dimensional decision matrix according to claim 1, characterized in that, The path optimization algorithm uses the ant colony algorithm, combined with replenishment task aggregation rules to realize delivery route planning. The task aggregation determination formula is as follows: In the formula, The straight-line distance between the two retail counters. and These are the geographical coordinates of the two retail counters. Set the radius for task aggregation.

8. The retail counter replenishment analysis method based on a multi-dimensional decision matrix according to claim 1, characterized in that, The user features of the service recipients are normalized to construct a user feature vector. The normalization formula is as follows: In the formula, Let i be the normalized value of the i-th user feature. For the original value of this feature for a single user, This is the minimum value of the feature. This is the maximum value of the feature; User feature vector is , These are normalized values ​​for age characteristics. Normalized values ​​for occupational characteristics To normalize the purchase frequency, This represents the normalized value of consumer preferences.

9. The retail counter replenishment analysis method based on a multi-dimensional decision matrix according to claim 1, characterized in that, The formula for updating the city's best-selling product data is: In the formula, This is an updated collection of best-selling products. This is a collection of best-selling products before the update. This is a collection of the latest top 10 best-selling products. These are the weighting coefficients.

10. A retail counter replenishment analysis method based on a multi-dimensional decision matrix, employing the retail counter replenishment analysis method based on a multi-dimensional decision matrix as described in any one of claims 1 to 9, characterized in that, include: The multi-dimensional decision matrix construction module is used to collect the location characteristics of retail counters, the user characteristics of service objects, and the local city's best-selling product data, and to construct a multi-dimensional decision matrix that integrates the location characteristics, the user characteristics of service objects, and the local city's best-selling product data; The candidate product set determination module is used to complete the compliance screening and recommendation of products based on the multi-dimensional decision matrix, and determine the candidate product set of the retail cabinet that is suitable for the current scenario and user needs; The candidate product display adjustment module is used to adjust the shelf display position of candidate products by combining their sales performance and attribute characteristics through dynamic allocation rules. The replenishment cycle analysis module is used to classify products into categories based on their historical sales frequency using a dynamic classification algorithm, and to formulate differentiated replenishment cycles for different product categories. The replenishment demand forecasting module is used to predict the replenishment demand of goods using a sales forecasting model and to plan the optimal delivery route using a route optimization algorithm.

Images