An erp intelligent management method applied to clothing e-commerce

Abstract

The application discloses an ERP intelligent management method applied to clothing e-commerce, and belongs to the technical field of ERP information management.The method filters out slow-selling products from local products based on a combination determination rule of an inventory yield ratio threshold value and a warehouse age threshold value, filters out corresponding similar competitive products for each slow-selling product based on product graphs and clothing multi-dimensional data, analyzes hot-selling competitive products based on the sales data of each similar competitive product in the corresponding merchant, and performs clustering on the prices of each hot-selling competitive product corresponding to the slow-selling product through a clustering algorithm, so that the optimal clustering cluster is obtained by taking the most quantity of price data points in the cluster as the optimal clustering cluster, and the cost of the corresponding slow-selling product and the highest price in the optimal clustering cluster are taken as a recommended pricing interval.The application effectively identifies and matches similar competitive products based on product graphs and clothing multi-dimensional data, and effectively identifies the price distribution of hot-selling products through a clustering algorithm, so as to provide a pricing adjustment strategy based on the price of the hot-selling competitive product for the slow-selling product.

Description

Technical Field

[0001] This invention relates to the field of ERP information management technology, and in particular to an intelligent ERP management method applied to apparel e-commerce. Background Technology

[0002] As a typical fashion retail sector, the apparel e-commerce industry is characterized by a wide variety of product styles, pronounced seasonality, and intense market competition. In the daily operations of apparel e-commerce companies, inventory management directly impacts their cash flow and operational efficiency. Research shows that apparel products have a short cycle from market launch to end of season. Once slow-moving stock appears, it leads to a series of problems such as inventory backlog, increased warehousing costs, and capital tied up, which can even severely impact normal business operations. Therefore, how to promptly identify slow-moving products and develop effective sales strategies has become a key issue for apparel e-commerce companies to improve operational efficiency. With the widespread application of ERP systems in the apparel e-commerce sector, ERP systems have been able to achieve unified management of product information, inventory data, and sales data. However, in the existing ERP system management technology, in terms of competitor analysis, it is still the operations staff who manually analyze and compare the data information stored in the ERP system. It is difficult to systematically consider multi-dimensional features such as product style, fabric, and color combination, resulting in inaccurate and incomplete matching of similar products, which affects subsequent pricing strategy adjustments. The invention patent with publication number CN118154273B discloses a method and system for intelligent competitive data analysis based on a data warehouse. It constructs a competitive analysis model, builds a target data warehouse using the various elements of the competitive analysis model, extracts target data from the target data warehouse and inputs it into the competitive analysis model to predict the sales data of competitors in the target period, obtaining predicted data; determines the rate of change of influencing factors based on the relationship between the predicted data and the actual data of competitors in the previous period; and adjusts product strategies according to the rate of change of influencing factors. The existing technologies described above have the following technical problems: they lack a competitor similarity assessment mechanism based on product attributes, fail to further identify and match similar competitors to our products, and therefore fail to fully consider the price distribution of best-selling products among similar competitors, thus failing to provide our slow-moving products with a pricing adjustment strategy based on the prices of best-selling competitors. Summary of the Invention

[0003] To address the technical problems existing in the prior art, this invention provides an ERP intelligent management method applied to apparel e-commerce, comprising the following steps: Product images, multi-dimensional apparel data, and sales data of each competitor and each local product are obtained from the competitor feature library and the local product feature library of the ERP system, respectively. The sales data includes price and sales volume. Based on a combination of the inventory-to-production ratio threshold and the inventory age threshold, slow-moving products are selected from local products. Based on product images and multidimensional apparel data, similar competitors are selected for each slow-moving product. Based on the sales data of each similar competitor among the merchants corresponding to similar competitors, determine whether the similar competitor is a best-selling competitor. The K-means clustering algorithm is used to cluster the prices of the best-selling competing products corresponding to the slow-moving products, resulting in final clusters. From each final cluster, the cluster with the highest price that is greater than the cost of the corresponding slow-moving product but less than the current price of the slow-moving product is selected as the first cluster. The first cluster with the most price data points is selected as the optimal cluster. The cost of the corresponding slow-moving product and the highest price in the optimal cluster are used as the recommended pricing range.

[0004] Furthermore, the multidimensional data of the clothing includes: style information, fabric information, and color information; The fabric information includes the fabric composition and the composition ratio of the components; The color information includes primary color and secondary color.

[0005] Furthermore, the combined judgment rule based on the inventory-to-production ratio threshold and the inventory age threshold, which filters out slow-moving products from local products, specifically involves: Obtain the production volume, current inventory level, and current inventory age of local products from the inventory records of the ERP system; If the ratio of the current inventory to the production of a local product is greater than or equal to a preset inventory-to-production ratio threshold, and the current inventory age is greater than or equal to a preset inventory age threshold, then the local product is considered a slow-moving product.

[0006] Furthermore, based on product images and multi-dimensional clothing data, the process of filtering out similar competitors from among the competing products for each slow-moving product specifically involves: Using a pre-trained deep convolutional neural network, feature vectors of slow-moving products and feature vectors of competing products are extracted from the product images of slow-moving products. Cosine similarity is used to calculate the product image similarity between the feature vectors of slow-moving products and the feature vectors of competing products. Based on multidimensional data analysis of clothing, the similarity of slow-moving products with each competitor in terms of style, fabric, and color is analyzed, and then information similarity is obtained through weighted fusion calculation. The overall similarity between slow-moving products and their competitors is calculated by weighted fusion of product image similarity and information similarity. Competitors whose product image similarity is greater than or equal to the first preset threshold, and / or whose information similarity is greater than or equal to the second preset threshold, and / or whose overall similarity is greater than or equal to the third preset threshold are considered as similar competitors to the corresponding slow-moving products.

[0007] Furthermore, the analysis of style similarity is specifically obtained by obtaining the ratio of the number of identical style information types between slow-selling products and competing products to the total number of style information types.

[0008] Furthermore, the fabric similarity is obtained through the following steps: Obtain the intersection and union of the fabric composition of slow-selling products and the fabric composition of competitors, and use the ratio of the intersection to the union as the composition similarity. The fabric components in the intersection are taken as the same components between the slow-moving products and the corresponding competitors. Based on the component composition ratio of the slow-moving products and the component composition ratio of the corresponding competitors, the percentage of each same component in the slow-moving products (H1) and the percentage of each same component in the corresponding competitors (H2) are obtained respectively. Based on the percentage of each identical ingredient in the slow-moving product, the identical ingredients are sorted in descending order to obtain the first ingredient sequence. Similarly, they are sorted in the corresponding competing products to obtain the second ingredient sequence. At the same time, each identical ingredient is assigned a component weight based on its percentage of content in the slow-moving product. The positions of the first component sequence and the second component sequence are compared to obtain the component sequence similarity. The similarity of the content of each identical component is obtained by calculating the content percentages H1 and H2 of each identical component. Then, the similarity of each component is weighted and fused according to the component weight of each identical component to obtain the comprehensive content similarity. The fabric similarity between slow-selling products and their corresponding competitors is calculated by weighted fusion of component similarity, component sequence similarity, and overall content similarity.

[0009] Furthermore, each identical ingredient is assigned a weight based on its percentage content in the slow-moving product, specifically as follows:

[0010] Let i be the component weight of the i-th common ingredient between the slow-moving product and its corresponding competitor. denoted as the percentage of the i-th identical ingredient in the slow-moving product, and N is the number of identical ingredients between the slow-moving product and its corresponding competitor. The step of comparing the positions of the first component sequence and the second component sequence to obtain the sequence similarity is as follows: the ratio of the number of identical components with the same position number in the first component sequence and the second component sequence to the total number of identical components is used as the sequence similarity. The similarity of the content of each identical component is obtained based on the content percentages H1 and H2 of each identical component. Specifically, the absolute value of the difference between the content percentages H1 and H2 of the identical component is subtracted from the numerical value as the corresponding component content similarity.

[0011] Furthermore, the color similarity is obtained through the following steps: Obtain the intersection and union of the primary colors of the slow-moving products and the primary colors of the competitors; and obtain the intersection and union of the secondary colors of the slow-moving products and the secondary colors of the competitors. The ratio of the intersection of primary colors to the union of primary colors is used as the primary color similarity, and the ratio of the intersection of secondary colors to the union of secondary colors is used as the secondary color similarity. The color similarity between slow-selling products and their corresponding competitors is obtained by weighted fusion of primary color similarity and secondary color similarity.

[0012] Furthermore, the determination of whether a similar competitor is a best-selling competitor based on the sales data of each similar competitor among the merchants corresponding to similar competitors specifically involves: Obtain the merchant ID corresponding to the similar competitor. Based on the merchant ID, query the similar competitors with the same merchant ID from all similar competitors corresponding to the slow-moving products of the similar competitor, and use them as the first similar competitor. The average sales volume and standard deviation of sales volume are calculated based on the sales volume of each first similar competitor. The product of the standard deviation of sales volume and the preset hot-selling threshold coefficient is summed with the average sales volume as the dynamic hot-selling threshold of the corresponding similar competitor. If the sales volume of similar competing products is greater than or equal to the preset minimum sales volume threshold for best-selling products, and is also greater than or equal to the corresponding dynamic best-selling threshold, then the similar competing products are considered best-selling competing products.

[0013] Furthermore, the K-means clustering algorithm is used to cluster the prices of each best-selling competitor corresponding to the slow-moving product, resulting in final clusters, specifically: The minimum and maximum prices of the corresponding best-selling competitors are obtained to form a price range. The price range is then divided into a preset number of equal parts. Each dividing node is used as an initial cluster center. Based on the initial cluster centers, the prices of the best-selling competitors are clustered using the K-means clustering algorithm to obtain multiple clusters. The mean of the price data in each cluster is calculated to obtain new cluster centers. Then, based on the new cluster centers, the prices of each best-selling competitor are clustered using the K-means clustering algorithm. This process is repeated until the preset number of iterations is reached to obtain the final clusters.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: By combining inventory production ratio and inventory age as a judgment rule, slow-moving products can be objectively and quantitatively screened from a large number of local products, avoiding reliance on human experience and improving the accuracy and efficiency of identification. By combining product images and multidimensional clothing data for similarity calculation, we can find truly similar competitors for slow-moving products from both visual and attribute perspectives. This overcomes the limitations of traditional methods that rely solely on text tags or manual comparison, and improves the comprehensiveness and accuracy of competitor matching. Based on the sales data of similar competing products from their respective merchants, combined with dynamic best-selling thresholds and minimum sales thresholds, it is possible to scientifically determine which similar competing products are best-selling products, thus avoiding the inclusion of competitors with occasional high or low sales in the analysis. The K-means clustering algorithm is used to perform unsupervised learning on the prices of best-selling competitors, automatically discovering the natural distribution pattern of prices, avoiding reliance on manual division of price tiers. From multiple first clusters, the cluster with the most data points is selected as the optimal cluster. A large number of data points indicates that the price range is used by the most best-selling competitors, effectively focusing on the mainstream price range and improving the market acceptance of pricing. By comparing the intersection / union of fabric components, percentage of component content, component sorting sequence, and content differences at multiple levels, it is possible to finely distinguish the similarity of fabrics in terms of component type, primary and secondary order, and proportion, thus significantly improving the matching accuracy of fabric dimensions. Attached Figure Description

[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of an ERP intelligent management method applied to apparel e-commerce according to the present invention; Figure 2 This is a flowchart of the screening process for similar competing products in the method of this invention; Figure 3 This is a flowchart of the process for obtaining fabric similarity in the method of the present invention; Figure 4 This is a structural block diagram of an ERP intelligent management system applied to the apparel e-commerce industry according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0020] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" and "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0021] Example 1 See Figures 1 to 3 As shown, the present invention provides an ERP intelligent management method for apparel e-commerce, which specifically includes the following steps: S1. Obtain product images, multi-dimensional apparel data, and sales data for each competitor and each local product from the competitor feature library and the local product feature library of the ERP system, respectively. S2. Based on the combined judgment rules of inventory-to-production ratio threshold and inventory age threshold, filter out slow-moving products from local products. S3. Based on product images and multi-dimensional apparel data, filter out similar competitors for each slow-moving product. S4. Based on the sales data of each similar competitor among the merchants corresponding to similar competitors, determine whether the similar competitor is a best-selling competitor. S5. Use the K-means clustering algorithm to cluster the prices of each best-selling competitor corresponding to the slow-moving product to obtain each final cluster. From each final cluster, select the cluster whose highest price is greater than the cost of the corresponding slow-moving product but less than the current price of the slow-moving product as the first cluster. Select the first cluster with the most price data points as the optimal cluster. S6. Use the cost of the corresponding slow-moving product and the highest price in the optimal cluster as the recommended pricing range.

[0022] S1. Obtain product images, multi-dimensional apparel data, and sales data for each competitor and each local product from the competitor feature library and the local product feature library of the ERP system, respectively.

[0023] In step S1, the multidimensional data of the clothing includes: style information, fabric information, and color information; The sales data includes price and sales volume.

[0024] Specifically, the style information includes, but is not limited to: clothing category, sleeve type, collar type, and pattern, etc.

[0025] Different product categories correspond to different style classification systems, such as T-shirts, shirts, dresses, coats, and jeans. Fits include slim fit, loose fit, straight fit, and flared fit.

[0026] Specifically, the fabric information includes the fabric composition and the proportions of the composition.

[0027] Specifically, the color information includes primary color and secondary color.

[0028] In step S1, the construction of the competitor feature database can be achieved by manually collecting data from the target collection channels or by using a distributed web crawler framework to perform targeted data crawling from the target collection channels. After obtaining the original competitor data, field cleaning is performed to remove invalid fields and outlier data. Then, the cleaned data is standardized in format, unifying the formats of different data sources into a structured data format. For example, the price field is standardized to RMB, the size field to the national standard code, and the color field to the standard color code, etc.

[0029] The distributed web crawler framework can be a customized distributed crawler framework based on the Scrapy framework, or it can be an asynchronous crawler framework based on Python. No restrictions are placed on the specific implementation of the crawler framework.

[0030] Specifically, the distributed web crawler framework includes a crawler scheduling module, a data retrieval module, and a data storage module: The crawler scheduling module is responsible for coordinating the allocation of collection tasks among multiple crawler nodes. The data crawling module is responsible for executing webpage requests, response parsing, and data extraction for the target link; The data storage module is responsible for storing the collected raw competitor data.

[0031] In step S1, the local product feature library is constructed by manually entering multi-dimensional clothing data of local products, which are the clothing products sold by the local clothing e-commerce platform.

[0032] The competitor feature library and the local product feature library can be stored using a MySQL relational database or a MongoDB non-relational database; no restrictions are placed on the database type.

[0033] S2. Based on the combined judgment rules of inventory-to-production ratio threshold and inventory age threshold, filter out slow-moving products from local products.

[0034] In step S2, the method of selecting slow-moving products from local products based on a combination of inventory-to-production ratio threshold and inventory age threshold specifically involves: Obtain the production volume, current inventory level, and current inventory age of local products from the inventory records of the ERP system; If the ratio of the current inventory to the production of a local product is greater than or equal to a preset inventory-to-production ratio threshold, and the current inventory age is greater than or equal to a preset inventory age threshold, then the local product is considered a slow-moving product.

[0035] The inventory-to-production ratio can effectively reflect the degree of unsold inventory: the higher the ratio, the less likely the produced goods are to be sold, and the greater the risk of unsold inventory; inventory age reflects the risk of unsold inventory over time. The longer the inventory age, the longer the products remain in the market without being cleared, and the more likely they are to become stockpiled inventory.

[0036] This solution identifies local products as slow-moving products if the ratio of current inventory to production is greater than or equal to a preset inventory-to-production ratio threshold, and the current inventory age is greater than or equal to a preset inventory age threshold. This avoids misjudgments caused by a single indicator (such as high inventory but short inventory age when seasonal products have just changed seasons), and eliminates interference from "false impressions of slow-moving new products" and "normal clearance stock".

[0037] S3. Based on product images and multi-dimensional apparel data, filter out similar competitors for each slow-moving product.

[0038] In step S3, based on product images and multidimensional clothing data, similar competitors are selected from among the competing products for each slow-moving product. Specifically: S31. Using a pre-trained deep convolutional neural network, extract the feature vectors of the slow-moving product images in the product images of slow-moving products and the feature vectors of the competitor images in the product images of each competitor; use cosine similarity to calculate the product image similarity between the feature vectors of the slow-moving product images and the feature vectors of each competitor image. S32. Based on multidimensional data analysis of clothing, analyze the similarity of style, fabric, and color between slow-moving products and competing products, and then calculate the information similarity through weighted fusion. S33. The overall similarity between slow-moving products and their competitors is obtained by weighted fusion calculation of product image similarity and information similarity. S34. Competitors whose product image similarity is greater than or equal to the first preset threshold, and / or whose information similarity is greater than or equal to the second preset threshold, and / or whose overall similarity is greater than or equal to the third preset threshold are considered as similar competitors of the corresponding slow-moving products.

[0039] This solution utilizes both visual and structured attributes for multimodal fusion comparison, which is more accurate than a single dimension. In step S32, the analysis of style similarity is specifically obtained by obtaining the ratio of the number of identical style information types between slow-moving products and competing products to the total number of style information types.

[0040] For example, style information includes four types: clothing category, sleeve type, collar type, and pattern. If a slow-moving product and a competitor's product have the same clothing category and sleeve type, but different collar type and pattern, then the number of the same style information types between the slow-moving product and the competitor's product is 2, and the style similarity is 2 / 4 = 0.5.

[0041] In step S32, the fabric similarity is obtained through the following steps: Sa321. Obtain the intersection and union of the fabric composition of slow-selling products and the fabric composition of competitors, and use the ratio of the intersection to the union as the composition similarity. Sa322. Take the fabric components in the intersection as the same components between the slow-moving products and the corresponding competitors, and obtain the percentage of each same component in the slow-moving products H1 and the percentage of each same component in the corresponding competitors H2 according to the component composition ratio of the slow-moving products and the component composition ratio of the corresponding competitors. Sa323. Based on the percentage of each identical ingredient in the slow-moving product, sort the identical ingredients in descending order to obtain the first ingredient sequence. Similarly, sort them in the corresponding competing products to obtain the second ingredient sequence. At the same time, assign ingredient weights to each identical ingredient based on the percentage of each identical ingredient in the slow-moving product. Sa324. The positions of the first component sequence and the second component sequence are compared to obtain the component sequence similarity. Sa325. Based on the content percentages H1 and H2 of each identical component, the similarity of the component content is obtained. Then, based on the component weights of each identical component, the similarity of each component is weighted and fused to obtain the comprehensive content similarity. Sa326. The fabric similarity between slow-selling products and corresponding competitors is obtained by weighted fusion calculation of component similarity, component sequence similarity and comprehensive content similarity.

[0042] This solution analyzes fabric similarity through a three-layer fusion structure: component similarity, component sequence similarity, and overall content similarity. The core of this approach is to perform a detailed, layer-by-layer modeling of the fabric, from "what components are there," to "how the components are ordered," and then to "what proportion each component accounts for," thereby comprehensively and accurately reflecting the true degree of similarity between two clothing products at the fabric level.

[0043] Composition similarity measures the degree of overlap in the types of fabric components (e.g., whether they all contain cotton, polyester fibers, etc.). Component sequence similarity measures whether the order of components is the same (e.g., cotton as the main component and polyester as the secondary component vs. polyester as the main component and cotton as the secondary component). Overall content similarity measures how close the specific percentages of each component are (e.g., 60% cotton vs 65% is more similar than 60% vs 30%).

[0044] In step Sa323, each identical component is assigned a component weight based on its percentage content in the slow-moving product, specifically as follows:

[0045] Let i be the component weight of the i-th common ingredient between the slow-moving product and its corresponding competitor. Let represent the percentage of the i-th identical ingredient in the slow-moving product, and N represent the number of identical ingredients between the slow-moving product and its corresponding competitor.

[0046] In step Sa324, the positions of the first component sequence and the second component sequence are compared to obtain the sequence similarity. Specifically, the ratio of the number of identical components with the same position number in the first component sequence and the second component sequence to the total number of identical components is used as the sequence similarity.

[0047] In step Sa325, the similarity of the content of each identical component is obtained based on the content percentages H1 and H2 of each identical component. Specifically, the absolute value of the difference between the content percentages H1 and H2 of the identical component is subtracted from the numerical value as the corresponding component content similarity.

[0048] In step S32, the color similarity is obtained through the following steps: Sb321. Obtain the intersection and union of the main colors of the main colors of the slow-moving products and the main colors of the competitors, and obtain the intersection and union of the secondary colors of the secondary colors of the slow-moving products and the secondary colors of the competitors. Sb322. The ratio of the intersection of primary colors to the union of primary colors is taken as the primary color similarity, and the ratio of the intersection of secondary colors to the union of secondary colors is taken as the secondary color similarity. Sb323: The color similarity between slow-moving products and corresponding competitors is obtained by weighted fusion of primary color similarity and secondary color similarity.

[0049] S4. Based on the sales data of each similar competitor among the merchants corresponding to similar competitors, determine whether the similar competitor is a best-selling competitor.

[0050] In step S4, based on the sales data of each similar competitor among the merchants corresponding to similar competitors, it is determined whether the similar competitor is a best-selling competitor, specifically as follows: S41. Obtain the merchant ID corresponding to the similar competitor, and query the similar competitor with the same merchant ID from all similar competitors corresponding to the slow-moving product corresponding to the similar competitor, and take it as the first similar competitor. S42. Calculate the average sales volume and standard deviation of sales volume based on the sales volume of each first similar competitor. Use the product of the standard deviation of sales volume and the preset hot-selling threshold coefficient and the sum of the average sales volume as the dynamic hot-selling threshold of the corresponding similar competitor. S43. If the sales volume of similar competing products is greater than or equal to the preset minimum sales volume threshold for best-selling products, and is also greater than or equal to the corresponding dynamic best-selling threshold, then the similar competing products are considered best-selling competing products.

[0051] In step S4 of this solution, the analysis unit is a merchant. Only similar competing products within the same merchant are compared, rather than directly comparing sales across merchants, thus eliminating the impact of differences in merchant size. A dynamic best-selling threshold is used, automatically calculated based on the merchant's own sales distribution. An absolute minimum sales threshold is also set as another line of defense to filter out occasional events with excessively low absolute sales. S5. Use the K-means clustering algorithm to cluster the prices of each best-selling competitor corresponding to the slow-moving product, and obtain each final cluster. From each final cluster, select the cluster whose highest price is greater than the cost of the corresponding slow-moving product but less than the current price of the slow-moving product as the first cluster. The first cluster with the most price data points in the cluster is selected as the optimal cluster.

[0052] In step S5, the K-means clustering algorithm clusters the prices of the best-selling competing products corresponding to the slow-moving products to obtain the final clusters, specifically: The minimum and maximum prices of the corresponding best-selling competitors are obtained to form a price range. The price range is then divided into a preset number of equal parts. Each dividing node is used as an initial cluster center. Based on the initial cluster centers, the prices of the best-selling competitors are clustered using the K-means clustering algorithm to obtain multiple clusters. The mean of the price data in each cluster is calculated to obtain new cluster centers. Then, based on the new cluster centers, the prices of each best-selling competitor are clustered using the K-means clustering algorithm. This process is repeated until the preset number of iterations is reached to obtain the final clusters.

[0053] Example 2 See Figure 4 As shown, the present invention also provides an ERP intelligent management system for apparel e-commerce, specifically including: The competitor feature library stores product images, multi-dimensional apparel data, and sales data for local products. A local product feature library stores product images, multi-dimensional apparel data, and sales data for each competitor; the sales data includes price and sales volume. The slow-moving product screening module uses a combination of judgment rules based on the inventory-to-production ratio threshold and the inventory age threshold to screen out slow-moving products from local products. The similar product filtering module, based on product images and multi-dimensional clothing data, filters out corresponding similar competitors for each slow-moving product. The best-selling product filtering module determines whether a product is a best-selling product based on the sales data of each similar competitor among the merchants corresponding to similar competitors. The pricing range recommendation module uses the K-means clustering algorithm to cluster the prices of each best-selling competitor corresponding to the slow-moving product, resulting in final clusters. From each final cluster, the cluster with the highest price greater than the cost of the corresponding slow-moving product but less than the current price of the slow-moving product is selected as the first cluster. The first cluster with the most price data points is selected as the optimal cluster. The cost of the corresponding slow-moving product and the highest price in the optimal cluster are used as the recommended pricing range.

[0054] Example 3 The present invention also provides an electronic device, including: a processor, a transmitting device, an input device, an output device, and a memory. The processor may be implemented using a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit, or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory may be implemented using a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM), and is used to store computer program code. The computer program code includes computer instructions. When the processor executes the computer instructions, the electronic device executes a method as described in any of the above possible implementation methods.

[0055] Example 4 The present invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor of an electronic device, cause the processor to perform a method as described in any of the above possible implementations.

[0056] The beneficial effects of this invention are as follows: By combining inventory production ratio and inventory age as a judgment rule, slow-moving products can be objectively and quantitatively screened from a large number of local products, avoiding reliance on human experience and improving the accuracy and efficiency of identification. By combining product images and multidimensional clothing data for similarity calculation, we can find truly similar competitors for slow-moving products from both visual and attribute perspectives. This overcomes the limitations of traditional methods that rely solely on text tags or manual comparison, and improves the comprehensiveness and accuracy of competitor matching. Based on the sales data of similar competing products from their respective merchants, combined with dynamic best-selling thresholds and minimum sales thresholds, it is possible to scientifically determine which similar competing products are best-selling products, thus avoiding the inclusion of competitors with occasional high or low sales in the analysis. The K-means clustering algorithm is used to perform unsupervised learning on the prices of best-selling competitors, automatically discovering the natural distribution pattern of prices, avoiding reliance on manual division of price tiers. From multiple first clusters, the cluster with the most data points is selected as the optimal cluster. A large number of data points indicates that the price range is used by the most best-selling competitors, effectively focusing on the mainstream price range and improving the market acceptance of pricing. By comparing the intersection / union of fabric components, percentage of component content, component sorting sequence, and content differences at multiple levels, it is possible to finely distinguish the similarity of fabrics in terms of component type, primary and secondary order, and proportion, thus significantly improving the matching accuracy of fabric dimensions.

[0057] In the description of this specification, the references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0058] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0059] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. An ERP intelligent management method applied to apparel e-commerce, characterized in that, Includes the following steps: Product images, multi-dimensional apparel data, and sales data of each competitor and each local product are obtained from the competitor feature library and the local product feature library of the ERP system, respectively. The sales data includes price and sales volume. Based on a combination of the inventory-to-production ratio threshold and the inventory age threshold, slow-moving products are selected from local products. Based on product images and multidimensional apparel data, similar competitors are selected for each slow-moving product. Based on the sales data of each similar competitor among the merchants corresponding to similar competitors, determine whether the similar competitor is a best-selling competitor. The K-means clustering algorithm is used to cluster the prices of the best-selling competing products corresponding to the slow-moving products, resulting in final clusters. From each final cluster, the cluster with the highest price that is greater than the cost of the corresponding slow-moving product but less than the current price of the slow-moving product is selected as the first cluster. The first cluster with the most price data points is selected as the optimal cluster. The cost of the corresponding slow-moving product and the highest price in the optimal cluster are used as the recommended pricing range.

2. The ERP intelligent management method for apparel e-commerce as described in claim 1, characterized in that, The multidimensional data of the clothing includes: style information, fabric information, and color information; The fabric information includes the fabric composition and the composition ratio of the components; The color information includes primary color and secondary color.

3. The ERP intelligent management method for apparel e-commerce as described in claim 2, characterized in that, The combined judgment rule based on the inventory-to-production ratio threshold and the inventory age threshold is used to filter out slow-moving products from local products, specifically as follows: Obtain the production volume, current inventory level, and current inventory age of local products from the inventory records of the ERP system; If the ratio of the current inventory to the production of a local product is greater than or equal to a preset inventory-to-production ratio threshold, and the current inventory age is greater than or equal to a preset inventory age threshold, then the local product is considered a slow-moving product.

4. The ERP intelligent management method for apparel e-commerce as described in claim 2, characterized in that, The process of filtering similar competitors from among competing products based on product images and multidimensional clothing data for each slow-moving product is as follows: Using a pre-trained deep convolutional neural network, feature vectors of slow-moving products and feature vectors of competing products are extracted from the product images of slow-moving products. Cosine similarity is used to calculate the product image similarity between the feature vectors of slow-moving products and the feature vectors of competing products. Based on multidimensional data analysis of clothing, the similarity of slow-moving products with each competitor in terms of style, fabric, and color is analyzed, and then information similarity is obtained through weighted fusion calculation. The overall similarity between slow-moving products and their competitors is calculated by weighted fusion of product image similarity and information similarity. Competitors whose product image similarity is greater than or equal to the first preset threshold, and / or whose information similarity is greater than or equal to the second preset threshold, and / or whose overall similarity is greater than or equal to the third preset threshold are considered as similar competitors to the corresponding slow-moving products.

5. The ERP intelligent management method for apparel e-commerce as described in claim 4, characterized in that, The analysis of style similarity is specifically obtained by obtaining the ratio of the number of identical style information types between slow-selling products and competing products to the total number of style information types.

6. The ERP intelligent management method for apparel e-commerce as described in claim 4, characterized in that, The fabric similarity is obtained through the following steps: Obtain the intersection and union of the fabric composition of slow-selling products and the fabric composition of competitors, and use the ratio of the intersection to the union as the composition similarity. The fabric components in the intersection are taken as the same components between the slow-moving products and the corresponding competitors. Based on the component composition ratio of the slow-moving products and the component composition ratio of the corresponding competitors, the percentage of each same component in the slow-moving products (H1) and the percentage of each same component in the corresponding competitors (H2) are obtained respectively. Based on the percentage of each identical ingredient in the slow-moving product, the identical ingredients are sorted in descending order to obtain the first ingredient sequence. Similarly, they are sorted in the corresponding competing products to obtain the second ingredient sequence. At the same time, each identical ingredient is assigned a component weight based on its percentage of content in the slow-moving product. The positions of the first component sequence and the second component sequence are compared to obtain the component sequence similarity. The similarity of the content of each identical component is obtained by calculating the content percentages H1 and H2 of each identical component. Then, the similarity of each component is weighted and fused according to the component weight of each identical component to obtain the comprehensive content similarity. The fabric similarity between slow-selling products and their corresponding competitors is calculated by weighted fusion of component similarity, component sequence similarity, and overall content similarity.

7. The ERP intelligent management method for apparel e-commerce as described in claim 6, characterized in that, Each identical ingredient is assigned a weight based on its percentage content in the slow-moving product, specifically as follows: ； Let i be the component weight of the i-th common ingredient between the slow-moving product and its corresponding competitor. denoted as the percentage of the i-th identical ingredient in the slow-moving product, and N is the number of identical ingredients between the slow-moving product and its corresponding competitor. The step of comparing the positions of the first component sequence and the second component sequence to obtain the sequence similarity is as follows: the ratio of the number of identical components with the same position number in the first component sequence and the second component sequence to the total number of identical components is used as the sequence similarity. The similarity of the content of each identical component is obtained based on the content percentages H1 and H2 of each identical component. Specifically, the absolute value of the difference between the content percentages H1 and H2 of the identical component is subtracted from the numerical value as the corresponding component content similarity.

8. The ERP intelligent management method for apparel e-commerce as described in claim 4, characterized in that, The color similarity is obtained through the following steps: Obtain the intersection and union of the primary colors of the slow-moving products and the primary colors of the competitors; and obtain the intersection and union of the secondary colors of the slow-moving products and the secondary colors of the competitors. The ratio of the intersection of primary colors to the union of primary colors is used as the primary color similarity, and the ratio of the intersection of secondary colors to the union of secondary colors is used as the secondary color similarity. The color similarity between slow-selling products and their corresponding competitors is obtained by weighted fusion of primary color similarity and secondary color similarity.

9. The ERP intelligent management method for apparel e-commerce as described in claim 1, characterized in that, The determination of whether a similar competitor is a best-selling competitor is based on the sales data of each similar competitor among the merchants corresponding to similar competitors, specifically: Obtain the merchant ID corresponding to the similar competitor. Based on the merchant ID, query the similar competitors with the same merchant ID from all similar competitors corresponding to the slow-moving products of the similar competitor, and use them as the first similar competitor. The average sales volume and standard deviation of sales volume are calculated based on the sales volume of each first similar competitor. The product of the standard deviation of sales volume and the preset hot-selling threshold coefficient is summed with the average sales volume as the dynamic hot-selling threshold of the corresponding similar competitor. If the sales volume of similar competing products is greater than or equal to the preset minimum sales volume threshold for best-selling products, and is also greater than or equal to the corresponding dynamic best-selling threshold, then the similar competing products are considered best-selling competing products.

10. The ERP intelligent management method applied to apparel e-commerce according to claim 1, characterized in that, The K-means clustering algorithm is used to cluster the prices of the best-selling competing products corresponding to the slow-moving products, resulting in the final clusters, as follows: The minimum and maximum prices of the corresponding best-selling competitors are obtained to form a price range. The price range is then divided into a preset number of equal parts. Each dividing node is used as an initial cluster center. Based on the initial cluster centers, the prices of the best-selling competitors are clustered using the K-means clustering algorithm to obtain multiple clusters. The mean of the price data in each cluster is calculated to obtain new cluster centers. Then, based on the new cluster centers, the prices of each best-selling competitor are clustered using the K-means clustering algorithm. This process is repeated until the preset number of iterations is reached to obtain the final clusters.

Images