Jewelry inventory management method and system combined with retail data analysis

Abstract

The invention discloses a jewelry inventory management method and system combined with retail data analysis, and relates to the technical field of inventory management, and the method comprises the steps: obtaining parameters and image information of jewelry inventory products, extracting feature vectors, carrying out clustering analysis, and forming a product category set and a typical portrait; collecting customer behavior data of the retail network, generating customer portraits, calculating the response probability of the customer portraits to product categories, and forming customer portrait distribution; based on customer portrait distribution and an optimization target, an inventory distribution scheme is generated by using an optimization algorithm; and executing inventory scheduling and updating the virtual inventory pool mapping relation. Therefore, the technical effect of improving the efficiency and accuracy of inventory management is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of inventory management, and particularly to a jewelry inventory management method and system combined with retail data analysis. Background Art

[0002] Traditional jewelry inventory management methods mainly rely on manual statistics and simple data analysis. By classifying and managing the basic parameters (such as material, weight, etc.) of jewelry products, inventory allocation and scheduling are achieved. At the same time, customer behavior data of retail outlets is usually used for market analysis, but the combination with inventory management is relatively limited. This management mode can meet the basic needs of enterprises to a certain extent, but when facing the complex SKUs of jewelry products and the diverse needs of customers, there are technical problems such as difficulty in achieving accurate inventory matching and efficient scheduling, resulting in inventory backlogs or out-of-stock phenomena. Summary of the Invention

[0003] The present invention provides a jewelry inventory management method and system combined with retail data analysis to solve the technical problems in the prior art that it is difficult to achieve accurate inventory matching and efficient scheduling, resulting in inventory backlogs or out-of-stock phenomena, and to achieve the technical effect of improving the efficiency and accuracy of inventory management.

[0004] In a first aspect, the present invention provides a jewelry inventory management method combined with retail data analysis, which includes: Obtain the product basic information of each model product in the jewelry inventory, where the product basic information includes parameter information and image information.

[0005] Perform feature engineering processing on the image information, and vectorize the feature engineering results and the parameter information respectively to obtain an image feature vector and a parameter feature vector.

[0006] Based on the image feature vector and the parameter feature vector, perform clustering analysis to obtain a product category set, where each is marked with a typical product category portrait and a product category space.

[0007] Obtain the customer behavior data of each retail outlet, generate multiple types of abstract customer portraits based on the customer behavior data, and establish a response probability matrix of each type of customer portrait to the product category, and output it as a customer portrait distribution.

[0008] According to the customer portrait distribution of each retail outlet, combined with a preset optimization goal, determine the inventory allocation plan of each retail outlet through an optimization algorithm.

[0009] Execute the inventory scheduling instruction based on the inventory allocation plan, and update the mapping relationship of the virtual inventory pool.

[0010] In a feasible implementation, obtain the product basic information of each model product in the jewelry inventory, including: Extract the parameter information, where the parameter information includes at least one of material type, design style label, and process complexity score.

[0011] Collect the image information, where the image information includes at least one of multi-angle high-definition product images, 3D model renderings, and AR try-on effect renderings.

[0012] Associate and store the parameter information and the image information with the product unique identification code.

[0013] In a feasible implementation, perform feature engineering processing on the image information, and vectorize the feature engineering results and the parameter information respectively to obtain an image feature vector and a parameter feature vector, including: Use a pre-trained convolutional neural network to extract features from the image information to generate the image feature vector.

[0014] Discriminate the parameter category of the parameter information, and combine with a preset parameter mapping rule to convert non-numerical parameters into numerical vectors.

[0015] Combine the numerical vector and the numerical parameter to generate the parameter feature vector.

[0016] In a feasible implementation, perform clustering analysis based on the image feature vector and the parameter feature vector to obtain a product category set, including: After standardizing the image feature vector and the parameter feature vector, perform feature vector splicing to generate a multi-modal feature vector.

[0017] Perform clustering analysis based on the multi-modal feature vector, and define the product category set according to the clustering result.

[0018] Analyze the clustering result to generate the typical product category portraits and the product category space of each product category.

[0019] In a feasible implementation, obtain the customer behavior data of each retail outlet, generate multiple types of abstract customer portraits based on the customer behavior data, and establish a response probability matrix of each type of customer portrait to the product category, and output it as a customer portrait distribution, including: Collect the customer behavior data of each retail outlet, including purchase records, browsing records, and try-on records.

[0020] Use data mining techniques to analyze the customer behavior data, extract customer features, and perform clustering analysis based on the customer features to generate multiple types of abstract customer portraits.

[0021] For each type of customer portrait, calculate the response probability for different product categories to form a response probability matrix.

[0022] Based on multiple types of abstract customer portraits, identify user portraits for the customer behavior data of each retail outlet, and generate the customer portrait distribution of each retail outlet.

[0023] In a feasible implementation manner, according to the customer portrait distribution of each retail outlet, combined with a preset optimization goal, determine the inventory allocation plan for each retail outlet through an optimization algorithm, further including: Define the inventory quantity of each product category allocated to each retail outlet as a decision variable.

[0024] Define the optimization goal, including a combination of maximizing the conversion rate and minimizing the scheduling cost. Among them, the conversion rate is calculated based on the customer portrait proportion and the response probability matrix, and the scheduling cost is calculated according to the transfer distance, logistics cost, and inventory holding cost.

[0025] Configure the constraint conditions, use the optimization algorithm to perform iterative optimization of the decision variables, and determine the inventory allocation plan for each retail outlet according to the preset weight coefficient. Among them, the constraint conditions include the minimum product placement quantity constraint for the outlet, the total product placement quantity constraint for a single item, and the scheduling time constraint.

[0026] In a feasible implementation manner, execute the inventory scheduling instruction based on the inventory allocation plan and update the mapping relationship of the virtual inventory pool, further including: Generate and execute the inventory scheduling instruction according to the inventory allocation plan, allocate the inventory from the virtual inventory pool to each retail outlet, and feedback the allocation log.

[0027] Update the mapping relationship of the virtual inventory pool according to the allocation log, and record the real-time status and distribution of the inventory.

[0028] In a second aspect, the present invention also provides a jewelry inventory management system combined with retail data analysis, which includes: A product basic information acquisition module, used to acquire the product basic information of each model product in the jewelry inventory, and the product basic information includes parameter information and image information.

[0029] A feature vector acquisition module, used to perform feature engineering processing on the image information, vectorize the feature engineering results and the parameter information respectively, and obtain an image feature vector and a parameter feature vector.

[0030] A product category set generation module, used to perform clustering analysis based on the image feature vector and the parameter feature vector to obtain a product category set, where each is marked with a typical product category portrait and a product category space.

[0031] A customer portrait distribution generation module, configured to obtain customer behavior data of each retail outlet, generate multiple types of abstract customer portraits based on the customer behavior data, establish a response probability matrix of each type of customer portrait to the product category, and output a customer portrait distribution.

[0032] An inventory allocation plan determination module, configured to determine an inventory allocation plan for each retail outlet according to the customer portrait distribution of each retail outlet, in combination with a preset optimization objective, through an optimization algorithm.

[0033] An inventory scheduling and updating module, configured to execute an inventory scheduling instruction based on the inventory allocation plan and update the mapping relationship of the virtual inventory pool.

[0034] The present invention discloses a jewelry inventory management method and system combining retail data analysis, including: obtaining parameter and image information of jewelry inventory products, extracting feature vectors and performing clustering analysis to form a product category set and a typical portrait; collecting customer behavior data of retail outlets, generating customer portraits and calculating their response probabilities to product categories to form a customer portrait distribution; generating an inventory allocation plan based on the customer portrait distribution and an optimization objective, through an optimization algorithm; executing inventory scheduling and updating the mapping relationship of the virtual inventory pool. The jewelry inventory management method and system combining retail data analysis disclosed by the present invention solve the technical problems of difficult accurate matching and efficient scheduling of inventory, resulting in inventory backlog or out-of-stock phenomena, and achieve the technical effect of improving the efficiency and accuracy of inventory management. Description of the Drawings

[0035] Figure 1 It is a flowchart of a jewelry inventory management method combining retail data analysis according to the present invention.

[0036] Figure 2 It is a structural diagram of a jewelry inventory management system combining retail data analysis according to the present invention.

[0037] Description of the reference numerals: a product basic information acquisition module 11, a feature vector acquisition module 12, a product category set generation module 13, a customer portrait distribution generation module 14, an inventory allocation plan determination module 15, an inventory scheduling and updating module 16. Detailed Embodiments

[0038] The above technical solutions will be described in detail below in conjunction with the accompanying drawings of the specification and specific embodiments to better understand the above technical solutions. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments only for explaining the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention. In addition, it should be noted that, for the sake of convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.

[0039] Embodiment 1, as Figure 1 is a schematic flowchart of a jewelry inventory management method combining retail data analysis according to the present invention, which includes: S100: Obtain the product basic information of each model product in the jewelry inventory, and the product basic information includes parameter information and image information.

[0040] Specifically, first interact with the current jewelry inventory database or jewelry inventory management system to extract the product basic information of existing jewelry products of multiple models (SKUs), so as to provide a basis for subsequent classification and analysis of jewelry products. Among them, the product basic information is divided into two categories: parameter information and image information. Parameter information usually covers factors such as the material of the jewelry (such as gold, platinum, diamond, etc.), weight, size, purity, and process complexity, which are key indicators for measuring the value of jewelry, applicable scenarios, and management costs; image information is an intuitive presentation of the appearance of the jewelry and an important part of the first impression that consumers have of the product, and is used to provide information about the design style, process details, and spatial three-dimensional sense of the jewelry.

[0041] In the above steps, the acquisition of image information effectively makes up for the problem of insufficient attention to the appearance characteristics of products in traditional inventory management. Based on the analysis and processing of image information, it is possible to more sensitively capture the market dynamics and consumer preferences of different styles of jewelry.

[0042] In some embodiments, obtaining the product basic information of each model product in the jewelry inventory includes: Extract the parameter information, where the parameter information includes at least one of material type, design style label, and process complexity score; collect the image information, where the image information includes at least one of multi-angle high-definition product images, 3D model renderings, and AR try-on effect diagrams; and associate and store the parameter information, the image information, and the product unique identification code.

[0043] Specifically, parameter information is a key element for understanding the characteristics of jewelry products, including material types (such as gold, platinum, diamonds, etc.), design style labels (such as minimalist style, retro style, court style, etc.), and process complexity scores (graded by professional jewelry appraisers or automated systems, indicating the difficulty and fineness of jewelry production), which provides a basic basis for the classification and management of jewelry.

[0044] Specifically, image information is an intuitive presentation of the appearance of jewelry, which is specifically divided into high-definition images of the product from multiple angles (showing the appearance details of the jewelry from different directions, such as the front, side, and top views of a ring), 3D model renderings (stereoscopic images of the jewelry generated using 3D modeling technology, which can be rotated 360° to view every detail of the jewelry), and AR try-on effect renderings (the wearing effect of the jewelry obtained through augmented reality technology).

[0045] Specifically, the product unique identification code is a unique code used to identify jewelry products. Associating the parameter information and image information with the unique identification code for storage is equivalent to creating a detailed electronic file for each piece of jewelry, facilitating precise management in the inventory management system. Exemplarily, the product unique identification code is SKU.

[0046] Specifically, the parameter information provides a quantitative description of the jewelry products for the system, facilitating the system to pre-classify the products according to factors such as material, style, and value, thereby providing structured data support for subsequent clustering analysis and ensuring the accuracy of mathematical operations. For example, by analyzing jewelry made of gold and platinum materials respectively, combined with their different processing techniques and market demands, more precise inventory layout and allocation can be achieved.

[0047] Specifically, the image information provides a basis for classification based on appearance features. For example, by analyzing features such as the gemstone inlay position and pattern complexity in the 3D model rendering, jewelry with similar design styles can be grouped together, realizing the clustering of jewelry from a visual perspective and making the inventory management more meticulous. In this way, the parameter information and the image information complement each other, providing the system with comprehensive product information input, covering the characteristics of jewelry products from the "insides" of material parameters to the "outsides" of appearance presentation, enabling subsequent classification, allocation, and scheduling operations to have sufficient data support.

[0048] S200: Perform feature engineering processing on the image information, and vectorize the feature engineering results and the parameter information separately to obtain an image feature vector and a parameter feature vector.

[0049] Specifically, feature engineering is the process of extracting effective features from raw data that are helpful for subsequent analysis and modeling. In this embodiment, the image information of the jewelry products contains a large amount of appearance details, which can be used to judge the styles, design elements, etc. of the jewelry. Therefore, it is necessary to perform feature engineering on the image information. In addition, the parameter information, including the material type, design style label, and process complexity score of the jewelry, also needs to be quantified so that it can be analyzed and modeled using mathematical methods.

[0050] Specifically, vectorization is the process of converting non-numerical features into numerical vectors, which helps with subsequent understanding and processing. In this embodiment, when processing image information, features of the image can be extracted through machine learning or deep learning methods. For example, a convolutional neural network can be used to identify features such as shapes, textures, and colors in the image. For parameter information, non-numerical features, such as material types (gold, platinum, diamond, etc.), need to be converted into numerical vectors so that they can be processed uniformly with the image feature vectors.

[0051] By obtaining the image feature vector and the parameter feature vector, the appearance and attribute information of the jewelry products can be understood and processed by the machine, which helps to quickly classify and cluster a large number of jewelry products, improving the accuracy and efficiency of inventory management.

[0052] In some embodiments, performing feature engineering on the image information, vectorizing the feature engineering results and the parameter information separately, and obtaining the image feature vector and the parameter feature vector includes: Using a pre-trained convolutional neural network to extract features from the image information to generate the image feature vector; performing parameter category discrimination on the parameter information, and combining a preset parameter mapping rule to convert non-numerical parameters into numerical vectors; combining the numerical vectors and the numerical parameters to generate the parameter feature vector.

[0053] Specifically, first, the pre-trained convolutional neural network learns features in the image, such as edges, textures, shapes, etc., through multiple convolutional and pooling operations. These learned features can be used to reflect key information such as the appearance design of the jewelry, the position and shape of gemstone inlays, etc., and generate corresponding image feature vectors. Among them, the image feature vector is presented in numerical form, and each dimension of the numerical value corresponds to a specific feature in the image to comprehensively describe the image content.

[0054] Specifically, for parameter information, non-numerical data such as material types (gold, platinum, diamond, etc.) and design style labels (simple style, retro style, court style, etc.) need to be converted into numerical vectors for mathematical calculation and analysis. The conversion process includes parameter category discrimination and mapping conversion based on preset parameter mapping rules. Through parameter category discrimination, parameters are classified into numerical parameters and non-numerical parameters; the preset parameter mapping rules define how to map various types of non-numerical parameters to the numerical space.

[0055] Exemplarily, non-numerical parameters are converted into continuous vector representations through pre-trained Embedding models (such as Word2Vec, GloVe) or by using an Embedding Layer in a deep learning model; then, for continuous parameters such as the weight or size of jewelry, standardization is performed to scale the data to the same range (mean of 0 and standard deviation of 1); then, the standardized results are binned according to a preset interval, dividing the continuous data into several intervals, where each interval corresponds to a numerical value, thereby discretizing the continuous parameters, and finally converting information such as material types and design styles into numerical values to facilitate comprehensive analysis together with the image feature vectors. For example, in subsequent clustering analysis, both the appearance and material of jewelry are considered, and products with similar appearance and material are grouped into one category, so as to more accurately identify the similarities and differences of jewelry products and provide more comprehensive information support for inventory allocation and management.

[0056] S300: Perform clustering analysis based on the image feature vector and the parameter feature vector to obtain a product category set, where each is marked with a typical product category portrait and a product category space.

[0057] Specifically, perform clustering analysis on multiple jewelry products based on the image feature vector, thereby dividing the jewelry products with diverse and complex models into multiple product categories with similar characteristics, forming a product category set; among them, the typical product category portrait is a description of the typical characteristics of the category, reflecting the appearance and attributes of representative products in the category; the product category space is the area occupied by each category in the feature vector space, defining the boundaries and scope of the category.

[0058] In some embodiments, performing clustering analysis based on the image feature vector and the parameter feature vector to obtain a product category set includes: After standardizing the image feature vector and the parameter feature vector, perform feature vector concatenation to generate a multi-modal feature vector; perform clustering analysis based on the multi-modal feature vector, and define a product category set according to the clustering result; analyze the clustering result to generate the typical product category portrait and the product category space of each product category.

[0059] Specifically, the image feature vector and the parameter feature vector respectively represent the appearance and attribute information of jewelry products, but their numerical ranges and units may be different. To ensure that these features can participate in the calculation fairly in subsequent analysis, they need to be standardized to ensure that each feature is within the same numerical range; for example, scale all feature values to between 0 and 1 or a distribution with a mean of 0 and a standard deviation of 1.

[0060] Specifically, combine the standardized image feature vector and parameter feature vector into a multi-dimensional vector as the multi-modal feature vector, which contains comprehensive information on the appearance and attributes of jewelry products.

[0061] Exemplarily, assume that the numerical range of the image feature vector is between -1 and 1, while the numerical range of the parameter feature vector is between 0 and 100. Then they need to be both converted to a distribution with a mean of 0 and a standard deviation of 1; then, concatenate the two standardized feature vectors to form a multi-modal feature vector with more dimensions. For example, if the image feature vector has 10 dimensions and the parameter feature vector has 5 dimensions, a 15-dimensional multi-modal feature vector is obtained after concatenation.

[0062] Specifically, use a clustering algorithm (such as K-means, DBSCAN, etc.) to perform clustering analysis on the multi-modal feature vector. Taking the K-means algorithm as an example, the number of clusters needs to be determined in advance, then randomly initialize the cluster centers, and then continuously iterate to assign each feature vector to the nearest cluster center and recalculate the positions of the cluster centers until the assignment of the feature vectors no longer changes or reaches the predetermined number of iterations.

[0063] Furthermore, define a product category set according to the clustering result, and generate a typical product category portrait and a product category space for each category. The typical product category portrait can be the average feature vector of each category or select the most representative feature vector in that category; the product category space is the region defined by the range of all feature vectors in each category, determined by statistical methods (such as calculating the maximum and minimum values). Exemplarily, the typical product category portrait is the multi-dimensional feature vector corresponding to the cluster center of each cluster in the clustering result.

[0064] Through the above steps, products with similar characteristics and market demands can be grouped into one category, which helps to more accurately evaluate the inventory levels and sales trends of each category. At the same time, it helps to gain a deeper understanding of market dynamics and consumer preferences.

[0065] S400: Obtain the customer behavior data of each retail outlet, generate multiple types of abstract customer portraits based on the customer behavior data, establish a response probability matrix of each type of customer portrait for the product categories, and output the customer portrait distribution.

[0066] Specifically, customer behavior data is the key information for understanding market demands and consumer preferences, which can reflect customers' interests and purchase intentions for different jewelry products, including purchase records, browsing records, try-on records, etc.

[0067] Through the analysis of customer behavior data, multiple types of abstract customer portraits can be generated. These customer portraits are general descriptions of customer group characteristics, and each type of customer portrait corresponds to different consumption habits and demand characteristics.

[0068] Specifically, the response probability matrix represents the purchase possibilities of different customer portraits for various product categories and is used to quantify the relationship between customer portraits and product categories; the customer portrait distribution is the data obtained by statistically analyzing the customer portraits of each retail outlet, which shows the proportion of different customer types in each retail outlet.

[0069] In some embodiments, obtaining the customer behavior data of each retail outlet, generating multiple types of abstract customer portraits based on the customer behavior data, and establishing a response probability matrix of each type of customer portrait for the product categories, and outputting the customer portrait distribution includes: Collect the customer behavior data of each retail outlet, including purchase records, browsing records, try-on records; use data mining techniques to analyze the customer behavior data, extract customer characteristics, and perform clustering analysis based on the customer characteristics to generate multiple types of abstract customer portraits; statistically calculate the response probabilities of each type of customer portrait for different product categories to form a response probability matrix; based on the multiple types of abstract customer portraits as the classification basis, perform user portrait recognition on the customer behavior data of each retail outlet to generate the customer portrait distribution of each retail outlet.

[0070] Specifically, first, customer behavior data of each retail outlet is collected through the sales system, e-commerce platform, and in-store fitting record system while complying with data privacy and protection. For example, the sales system can record customers' purchase records, including the product categories purchased, purchase amounts, and purchase times; the e-commerce platform can track customers' browsing records, including the product pages browsed and the stay times. In-store fitting records of customers can be recorded through fitting devices or manually. Then, data mining techniques are used to analyze the customer behavior data to extract customer characteristics, including steps such as data cleaning, data transformation, and feature extraction, to convert the raw data into a format suitable for modeling. Among them, customer characteristics can include purchase preferences, consumption levels, shopping frequencies, preferred product categories, etc. Next, based on the above method steps for clustering analysis of jewelry products, clustering analysis is performed on the customer characteristics to divide customers into different groups. For example, according to the purchase amount and purchase frequency, customers can be divided into high-value customers, medium-value customers, and low-value customers. Each type of customer profile has its unique characteristics. For example, high-value customers may pay more attention to the brand and design of jewelry, while low-value customers may focus more on price and practicality.

[0071] Furthermore, after the clustering analysis, a response probability matrix of each type of customer profile for different product categories is established according to the clustering results. Among them, the rows of the matrix represent different customer profiles, and the columns of the matrix represent different product categories. Each cell stores the corresponding response probability. Exemplarily, through historical data, the response probabilities of each type of customer profile for purchasing different product categories are statistically analyzed. For example, if the high-value customer group purchased 100 jewelry products in the past year, 50 of which belong to the diamond jewelry category, 30 belong to the gold jewelry category, and 20 belong to the silver jewelry category, then the response probability of high-value customers for diamond jewelry can be calculated as 50%.

[0072] Finally, based on multiple types of abstract customer profiles as the basis for division, user portrait recognition is performed on the customer behavior data of retail outlets, and the proportions of different customer groups in each retail outlet are statistically analyzed to generate the customer portrait distribution of each retail outlet.

[0073] Through the above steps, understanding the customer portrait distribution of different retail outlets and the response probabilities of each type of customer profile for product categories provides a more accurate basis for inventory allocation, can reduce inventory costs, improve inventory turnover rates, and at the same time helps to enhance the shopping experience of customers.

[0074] S500: According to the customer portrait distribution of each retail outlet, combined with a preset optimization goal, determine the inventory allocation plan for each retail outlet through an optimization algorithm.

[0075] In the entire inventory management solution, the role of this step is to achieve precise inventory allocation. By combining the customer profile distribution and optimization goals, a more reasonable inventory allocation plan can be formulated, improving the utilization efficiency of inventory and the sales success rate.

[0076] For example, if the customer profile of a retail outlet shows a high demand for a certain type of product and the inventory at that outlet is insufficient, the optimization algorithm will automatically increase the inventory allocation to that outlet to meet market demand. At the same time, by optimizing the scheduling cost, the enterprise can reduce operating costs and improve profitability.

[0077] In some embodiments, according to the customer profile distribution of each retail outlet, combined with a preset optimization goal, the inventory allocation plan for each retail outlet is determined through an optimization algorithm, and further includes: Defining the inventory quantity of each product category allocated to each retail outlet as a decision variable; defining optimization goals, including a combination of maximizing the conversion rate and minimizing the scheduling cost, where the conversion rate is calculated based on the customer profile proportion and the response probability matrix, and the scheduling cost is calculated according to the transfer distance, logistics cost, and inventory holding cost; configuring constraint conditions, using the optimization algorithm to perform iterative optimization of the decision variable, and determining the inventory allocation plan for each retail outlet according to a preset weight coefficient, where the constraint conditions include the minimum stocking quantity constraint for the outlet, the total stocking quantity constraint for the single product, and the scheduling time constraint.

[0078] Specifically, the optimization goal is the core guiding principle for inventory allocation, including maximizing the conversion rate and minimizing the scheduling cost. Among them, the conversion rate refers to the proportion of transactions successfully completed within a certain period of time, which can be calculated through the proportion of the customer profile and the corresponding response probability matrix; the scheduling cost refers to various costs generated during the inventory allocation process, including the transfer distance, logistics cost, and inventory holding cost.

[0079] Specifically, the decision variable refers to the variable that needs to be determined during the optimization process. In this embodiment, it is the inventory quantity of each product category allocated to each retail outlet; the constraint conditions refer to the limiting conditions that need to be met during the optimization process, including the minimum stocking quantity constraint for the outlet (the minimum inventory level that each outlet must maintain), the total stocking quantity constraint for the single product (the total inventory quantity of each product category), and the scheduling time constraint (the maximum time allowed for inventory transfer).

[0080] Specifically, first, define the inventory quantity of each product category allocated to each retail outlet as a decision variable. For example, assume that a jewelry brand has three product categories: diamond jewelry, gold jewelry, and silver jewelry, and each category needs to be allocated to different retail outlets. Exemplarily, the decision variable can be represented as a matrix, where the rows represent the product categories, the columns represent the retail outlets, and each element in the matrix represents the inventory quantity of the product category at that retail outlet.

[0081] Specifically, next, an optimization objective is defined, which is characterized as a multi-objective function, including maximizing the conversion rate and minimizing the scheduling cost. Among them, the conversion rate is calculated through the proportion of customer portraits and the response probability matrix; the scheduling cost needs to consider the transfer distance, logistics cost, and inventory holding cost. For example, if the distance for transferring inventory from one outlet to another is far, the logistics cost will increase; if the inventory holding time is long, the holding cost will also increase. Then, configuration constraints are set, which include the minimum inventory allocation constraint for outlets, the total inventory allocation constraint for single products, and the scheduling time constraint. For example, each outlet must maintain a certain amount of inventory to meet daily sales needs, which is the minimum inventory allocation constraint for outlets; the total inventory of each product category cannot exceed the production capacity and supply chain capacity of the brand side, which is the total inventory allocation constraint for single products; inventory transfer needs to be completed within a certain time, which is the scheduling time constraint.

[0082] Furthermore, an optimization algorithm is used to iteratively optimize the decision variables. Among them, according to the preset weight coefficients, the importance of different optimization objectives is weighed to find the optimal inventory allocation plan. For example, if the brand side pays more attention to the conversion rate, then improving the conversion rate is given priority in the optimization (increasing the weight of the conversion rate in the multi-objective function), even if the scheduling cost increases slightly; correspondingly, if the brand side pays more attention to cost control, then the optimization algorithm will give priority to reducing the scheduling cost, even if the conversion rate decreases slightly.

[0083] Exemplarily, an optimization algorithm is used to solve the inventory allocation plan: First, a random initial inventory allocation plan is generated; then, it is checked whether the initial inventory allocation plan meets the constraints such as minimum inventory, total inventory, and scheduling time, and the objective value (the value of the multi-objective function) of the current plan is calculated based on the conversion rate and scheduling cost; then, methods such as gradient descent and simulated annealing are used to gradually optimize the decision variables, and after the calculation converges, the optimal inventory allocation plan is output.

[0084] Through the above optimization objective of combining maximizing the conversion rate and minimizing the scheduling cost based on the customer portrait distribution, the optimal inventory allocation plan is solved through an optimization algorithm, and it is ensured that the constraints such as inventory quantity and scheduling time are met, so as to achieve efficient inventory management, improve the profitability of retail outlets, and at the same time, help customers more easily purchase the products they are interested in, thereby improving customer satisfaction and loyalty.

[0085] S600: Execute the inventory scheduling instruction based on the inventory allocation plan and update the mapping relationship of the virtual inventory pool.

[0086] Specifically, the inventory allocation plan determines the inventory quantity to be allocated to each retail outlet, including the quantity of each type of jewelry product allocated to each outlet; the inventory scheduling instruction is a specific execution carrier generated based on the inventory allocation plan for guiding the actual transfer and allocation of inventory.

[0087] In some embodiments, based on the inventory allocation plan, executing the inventory scheduling instruction and updating the mapping relationship of the virtual inventory pool further includes: Generating and executing an inventory scheduling instruction according to the inventory allocation plan, allocating the inventory from the virtual inventory pool to each retail outlet, and feeding back the allocation log; updating the mapping relationship of the virtual inventory pool according to the allocation log to record the real-time status and distribution of the inventory.

[0088] Specifically, first, generate an inventory scheduling instruction according to the inventory allocation plan to clarify the product categories and quantities that each retail outlet needs to receive; then, execute the inventory scheduling instruction, allocate the inventory from the virtual inventory pool to each retail outlet, and perform the actual inventory transfer and transportation process to ensure that the products can reach the designated retail outlets on time and in the required quantities; next, update the mapping relationship of the virtual inventory pool according to the results of the scheduling instruction. For example, when a batch of diamond jewelry is transferred from the virtual inventory pool to a certain retail outlet, the quantity of the diamond jewelry in the virtual inventory pool will decrease accordingly, while the inventory quantity of that retail outlet will increase. By updating the mapping relationship of the virtual inventory pool, it can be ensured that the data in the inventory management system is consistent with the actual inventory and reflects the dynamic changes of the inventory in real time.

[0089] In summary, a jewelry inventory management method combining retail data analysis provided by the present invention has the following technical effects: By obtaining the parameter and image information of jewelry inventory products, extracting feature vectors and performing clustering analysis to form a product category set and a typical portrait; collecting customer behavior data of retail outlets, generating customer portraits and calculating their response probabilities to product categories to form a customer portrait distribution; generating an inventory allocation plan based on the customer portrait distribution and the optimization goal; executing inventory scheduling and updating the mapping relationship of the virtual inventory pool, thereby achieving the technical effects of improving the efficiency and accuracy of inventory management.

[0090] Embodiment 2, as Figure 2 is a schematic structural diagram of a jewelry inventory management system combining retail data analysis according to the present invention. For example, Figure 1 in the present invention, the flow schematic diagram of a jewelry inventory management method can be implemented through a structure such as Figure 2 shown.

[0091] Based on the same concept as the jewelry inventory management method combined with retail data analysis in the above embodiments, the present invention also provides a jewelry inventory management system combined with retail data analysis, including: A product basic information acquisition module 11, configured to acquire the product basic information of each model product in the jewelry inventory, where the product basic information includes parameter information and image information.

[0092] A feature vector acquisition module 12, configured to perform feature engineering processing on the image information, vectorize the feature engineering results and the parameter information respectively, and acquire an image feature vector and a parameter feature vector.

[0093] A product category set generation module 13, configured to perform clustering analysis based on the image feature vector and the parameter feature vector to obtain a product category set, where each is correspondingly marked with a typical product category portrait and a product category space.

[0094] A customer portrait distribution generation module 14, configured to acquire the customer behavior data of each retail outlet, generate multiple types of abstract customer portraits based on the customer behavior data, and establish a response probability matrix of each type of customer portrait to the product category, and output a customer portrait distribution.

[0095] An inventory allocation scheme determination module 15, configured to determine the inventory allocation scheme of each retail outlet according to the customer portrait distribution of each retail outlet, in combination with a preset optimization objective, through an optimization algorithm.

[0096] An inventory scheduling and updating module 16, configured to execute an inventory scheduling instruction based on the inventory allocation scheme and update the mapping relationship of the virtual inventory pool.

[0097] In some embodiments, the product basic information acquisition module 11 includes: A parameter information extraction unit, configured to extract the parameter information, where the parameter information includes at least one of a material type, a design style label, and a process complexity score.

[0098] An image information acquisition unit, configured to acquire the image information, where the image information includes at least one of multi-angle high-definition product images, 3D model rendering diagrams, and AR try-on effect diagrams.

[0099] An associated storage unit, configured to store the parameter information, the image information, and the product unique identification code in an associated manner.

[0100] In some embodiments, the feature vector acquisition module 12 includes: An image feature extraction unit, configured to extract features from the image information by using a pre-trained convolutional neural network to generate the image feature vector.

[0101] A parameter category discrimination and numerical conversion unit is used to discriminate the parameter category of the parameter information and convert non-numerical parameters into numerical vectors in combination with a preset parameter mapping rule.

[0102] A parameter feature vector generation unit is used to generate the parameter feature vector in combination with the numerical vector and the numerical parameter.

[0103] In some embodiments, the product category set generation module 13 includes: A multi-modal feature vector generation unit is used to splice the feature vectors after normalizing the image feature vector and the parameter feature vector to generate a multi-modal feature vector.

[0104] A product category set definition unit is used to perform clustering analysis based on the multi-modal feature vector and define a product category set according to the clustering result.

[0105] A typical product category portrait and product category space generation unit is used to analyze the clustering result and generate the typical product category portrait and the product category space of each product category.

[0106] In some embodiments, the customer portrait distribution generation module 14 includes: A customer behavior data collection unit is used to collect customer behavior data of each retail outlet, including purchase records, browsing records, and try-on records.

[0107] A customer portrait generation unit is used to analyze the customer behavior data using data mining techniques, extract customer features, and perform clustering analysis based on the customer features to generate multiple types of abstract customer portraits.

[0108] A response probability matrix generation unit is used to statistically calculate the response probability of each type of customer portrait to different product categories to form a response probability matrix.

[0109] A customer portrait distribution generation unit is used to perform user portrait recognition on the customer behavior data of each retail outlet based on multiple types of abstract customer portraits as the division basis, and generate the customer portrait distribution of each retail outlet.

[0110] In some embodiments, the inventory allocation scheme determination module 15 includes: A decision variable definition unit is used to define the inventory quantity of each product category allocated to each retail outlet as a decision variable.

[0111] An optimization objective definition unit is used to define an optimization objective, including a combination of maximizing the conversion rate and minimizing the scheduling cost, where the conversion rate is calculated based on the customer portrait proportion and the response probability matrix, and the scheduling cost is calculated according to the transfer distance, logistics cost, and inventory holding cost.

[0112] A constraint configuration and optimization unit is used to configure constraints, perform iterative optimization of the decision variables using an optimization algorithm, and determine the inventory allocation plan for each retail outlet according to a preset weight coefficient. Among them, the constraints include the minimum stocking quantity constraint for outlets, the total stocking quantity constraint for single items, and the scheduling time constraint.

[0113] In some embodiments, the inventory scheduling and update module 16 further includes: An inventory scheduling and allocation unit is used to generate and execute an inventory scheduling instruction according to the inventory allocation plan, allocate the inventory from the virtual inventory pool to each retail outlet, and feedback the allocation log.

[0114] A virtual inventory pool update unit is used to update the mapping relationship of the virtual inventory pool according to the allocation log, and record the real-time status and distribution of the inventory.

[0115] It should be understood that the embodiments mentioned in this specification focus on their differences from other embodiments. The specific embodiments in the foregoing Embodiment 1 are equally applicable to the jewelry inventory management system combining retail data analysis described in Embodiment 2. For the sake of brevity of the specification, no further elaboration is made here.

[0116] It should be understood that the disclosed embodiments of the present invention and the above descriptions can enable those skilled in the art to implement the present invention using the present invention. At the same time, the present invention is not limited to the above-mentioned part of the embodiments. It should be understood that those of ordinary skill in the art can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A jewelry inventory management method combined with retail data analysis, characterized in that: include: Obtaining basic product information of each model product in the jewelry inventory, wherein the basic product information includes parameter information and image information; Performing feature engineering processing on the image information, and respectively quantizing the feature engineering results and the parameter information to obtain an image feature vector and a parameter feature vector; Performing cluster analysis based on the image feature vector and the parameter feature vector to obtain a product category set, wherein each corresponding label is marked with a typical product category portrait and a product category space; Obtain customer behavior data of each retail outlet, generate multiple types of abstract customer portraits based on the customer behavior data, and establish a response probability matrix of each type of customer portrait to the product category, and output the customer portrait distribution; According to the customer profile distribution of each retail outlet, combined with the preset optimization goal, the inventory allocation plan of each retail outlet is determined through the optimization algorithm; The inventory scheduling instruction is executed based on the inventory allocation plan, and the mapping relationship of the virtual inventory pool is updated.

2. A jewelry inventory management method combined with retail data analysis as claimed in claim 1, characterized in that: Get basic product information for each model in your jewelry inventory, including: Extracting the parameter information, wherein the parameter information includes at least one of a material type, a design style label, and a process complexity score; Collecting the image information, wherein the image information includes at least one of a multi-angle product high-definition image, a 3D model rendering, and an AR trial wearing effect image; The parameter information, the image information and the product unique identification code are associated and stored.

3. A jewelry inventory management method combined with retail data analysis as claimed in claim 2, characterized in that: Performing feature engineering processing on the image information, and respectively vectorizing the feature engineering results and the parameter information to obtain an image feature vector and a parameter feature vector, including: Using a pre-trained convolutional neural network to extract features from the image information to generate the image feature vector; Perform parameter category discrimination on the parameter information, and convert the non-numeric parameters into numeric vectors in combination with preset parameter mapping rules; The numerical vector and the numerical parameter are combined to generate the parameter feature vector.

4. A jewelry inventory management method combined with retail data analysis as claimed in claim 3, characterized in that: Based on the image feature vector and the parameter feature vector, cluster analysis is performed to obtain a product category set, including: After normalizing the image feature vector and the parameter feature vector, feature vector splicing is performed to generate a multimodal feature vector; Performing cluster analysis based on the multimodal feature vectors, and defining a product category set according to the clustering results; Analyze the clustering results and generate the typical product category portrait and the product category space for each product category.

5. A jewelry inventory management method combined with retail data analysis as claimed in claim 4, characterized in that: Obtain customer behavior data from each retail outlet, generate multiple types of abstract customer portraits based on the customer behavior data, and establish a response probability matrix for each type of customer portrait to the product category, and output the customer portrait distribution, including: Collect customer behavior data from various retail outlets, including purchase records, browsing records, and trial records; Using data mining technology to analyze the customer behavior data, extract customer characteristics, and perform cluster analysis based on the customer characteristics to generate multiple types of abstract customer portraits; For each type of customer portrait, calculate the probability of their response to different product categories to form a response probability matrix; Based on multiple types of abstract customer portraits as the basis for classification, user portrait recognition is performed on the customer behavior data of each retail outlet to generate the customer portrait distribution of each retail outlet.

6. A jewelry inventory management method combined with retail data analysis as claimed in claim 5, characterized in that: According to the customer profile distribution of each retail outlet, combined with the preset optimization goal, the inventory allocation plan of each retail outlet is determined through the optimization algorithm, which also includes: Define the inventory quantity of each product category allocated to each retail outlet as the decision variable; Define optimization goals, including a combination of maximizing the transaction rate and minimizing the dispatch cost. The transaction rate is calculated based on the customer profile ratio and the response probability matrix, and the dispatch cost is calculated based on the transfer distance, logistics costs, and inventory holding costs. Configure constraints, use an optimization algorithm to iteratively optimize the decision variables, and determine the inventory allocation plan for each retail outlet based on preset weight coefficients, wherein the constraints include the minimum distribution quantity constraint of the outlet, the total distribution quantity constraint of a single product, and the scheduling time constraint.

7. A jewelry inventory management method combined with retail data analysis as claimed in claim 6, characterized in that: Executing the inventory scheduling instruction based on the inventory allocation plan and updating the mapping relationship of the virtual inventory pool also includes: Generate and execute inventory dispatch instructions according to the inventory allocation plan, allocate inventory from the virtual inventory pool to each retail outlet, and provide feedback on the allocation log; The mapping relationship of the virtual inventory pool is updated according to the allocation log, and the real-time status and distribution of the inventory are recorded.

8. A jewelry inventory management system combined with retail data analysis, characterized in that: The system is used to execute the jewelry inventory management method combined with retail data analysis as described in any one of claims 1 to 7, comprising: A product basic information acquisition module, used to acquire the product basic information of each model of the jewelry inventory, wherein the product basic information includes parameter information and image information; A feature vector acquisition module is used to perform feature engineering processing on the image information, and to quantize the feature engineering results and the parameter information respectively, to obtain an image feature vector and a parameter feature vector; A product category set generation module, configured to perform cluster analysis based on the image feature vector and the parameter feature vector to obtain a product category set, wherein each corresponding label is provided with a typical product category portrait and a product category space; A customer profile distribution generation module is used to obtain customer behavior data of each retail outlet, generate multiple types of abstract customer profiles based on the customer behavior data, and establish a response probability matrix of each type of customer profile to the product category, and output the customer profile distribution; An inventory allocation plan determination module is used to determine the inventory allocation plan of each retail outlet through an optimization algorithm based on the customer profile distribution of each retail outlet and a preset optimization goal; The inventory scheduling and updating module is used to execute the inventory scheduling instruction based on the inventory allocation plan and update the mapping relationship of the virtual inventory pool.

Images