Retail inventory optimization method and system based on full-link data

Abstract

The invention discloses a retail inventory optimization method and system based on full-link data, and relates to the technical field of inventory management, and the method comprises the steps: building a full-link data collection channel, cooperatively collecting the multi-dimensional data of the whole process of a retail business, and carrying out the layering preprocessing of the data, and constructing a full-link data resource pool; meanwhile, an inventory association analysis model is constructed, and association logic of the multi-dimensional data, the inventory state and the on-time commodity classification is defined; based on a model output result, generating an initial inventory optimization scheme including commodity replenishment, inventory allocation, unsalable commodity processing, in-time commodity grading processing and other schemes; and a dynamic adjustment and supply chain closed-loop feedback mechanism is established, the scheme is adjusted in real time, and the disposal data is fed back to the upstream supply chain, so that the problems of insufficient data integration, lack of a standardized system for the in-time commodity disposal, lack of supply chain collaboration and the like in the traditional technology are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of inventory management technology, specifically to a retail inventory optimization method and system based on end-to-end data. Background Technology

[0002] In today's booming retail industry, the scale and complexity of retail operations are increasing daily. Retail inventory management, as a crucial link in retail operations, plays a vital role in ensuring stable product supply, reducing operating costs, and improving customer satisfaction. With the rapid development of information technology, the data dimensions involved in retail operations are becoming increasingly rich, covering multiple aspects such as supply chain, sales, logistics, users, and product attributes. Fully mining and utilizing this data will optimize retail inventory management, thereby improving the overall operational efficiency and competitiveness of the retail business. Therefore, how to build a scientific, efficient, and comprehensive retail inventory management technology solution has become a crucial issue that the retail industry urgently needs to address.

[0003] Traditional retail inventory management methods have revealed numerous drawbacks when addressing the complex demands of modern retail operations. On one hand, traditional methods rely heavily on single-dimensional data statistics or manual experience for decision-making, lacking a mechanism for integrating data across the entire supply chain. This prevents inventory decisions from comprehensively and accurately reflecting the true state of retail operations, resulting in a lack of scientific and systematic inventory management. On the other hand, traditional technologies fail to establish standardized hierarchical systems and disposal procedures for near-expiry goods. The disposal of near-expiry goods often lacks clear standards, leading to arbitrary handling methods and an inability to accurately address the actual condition of the goods. This not only wastes resources but may also negatively impact the company's economic benefits and brand image. Furthermore, traditional technologies are significantly inadequate in supply chain collaboration. Data on near-expiry goods disposal fails to effectively link with the upstream supply chain, preventing upstream procurement from adjusting purchasing strategies in a timely manner based on actual inventory levels. This further exacerbates the irrationality of inventory management and makes it difficult to adapt to the dynamic changes throughout the entire retail process. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies by providing a retail inventory optimization method and system based on end-to-end data. This involves establishing an end-to-end data acquisition channel to collaboratively collect multi-dimensional data from the entire retail business process, and performing layered preprocessing to construct an end-to-end data resource pool. Simultaneously, an inventory correlation analysis model is built to clarify the correlation logic between multi-dimensional data and inventory status and the classification of near-expiry goods. Based on the model output, an initial inventory optimization plan is generated, including solutions for replenishment, inventory allocation, handling of slow-moving goods, and graded disposal of near-expiry goods. Furthermore, a dynamic adjustment and supply chain closed-loop feedback mechanism is established to adjust the plan in real time and feed the disposal data back to the upstream supply chain. This effectively solves problems such as insufficient data integration, lack of standardization in the disposal of near-expiry goods, and lack of supply chain collaboration in traditional technologies.

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies. On the one hand, it provides a retail inventory optimization method based on end-to-end data, which includes the following specific steps:

[0006] S1: Establish a full-chain data collection channel to collaboratively collect data from the supply chain, sales, logistics, user, and product attribute categories according to preset rules;

[0007] S2: Perform hierarchical preprocessing on the collected multi-dimensional data across the entire data chain to build a full-chain data resource pool based on a master-slave architecture distributed database;

[0008] S3: Construct an inventory correlation analysis model based on the multiple linear regression algorithm, determine the correlation weights of variables through model training, and form standardized correlation analysis rules;

[0009] S4: Based on the output of the inventory correlation analysis model, combined with the preset inventory safety threshold and near-expiry classification threshold, an initial inventory optimization plan is generated;

[0010] S5: Real-time collection of dynamic data on retail operations and execution data on the disposal of near-expiry goods, enabling dynamic adjustment of initial inventory optimization plans, closed-loop feedback of the supply chain, and final distribution of the final plan to the end user.

[0011] Furthermore, in step S1, the end-to-end data acquisition channel includes five types of interfaces: a RESTful interface that interfaces with the supplier management system to pull supply chain data at a preset period; a data acquisition interface for store sales terminals to collect sales data based on the TCP / IP protocol; a WebService interface that interfaces with the logistics management system to obtain logistics data in real time; a crawler interface for user data to collect relevant user data; and a JDBC interface that interfaces with the product information management system to extract product attribute data in batches.

[0012] Furthermore, in step S1, the multi-dimensional data includes supply chain data, sales data, logistics data, user data, and product attribute data. Supply chain data includes supplier capacity data, raw material supply cycle data, and supply batch restriction data. Sales data includes real-time sales data of each store, product inventory balance data, and promotional activity data. Logistics data includes warehousing and sorting timeliness data, cross-regional transportation cycle data, and logistics capacity data. User data includes user consumption preference data, repurchase frequency data, and regional demand distribution data. Product attribute data includes product shelf life data, historical loss rate data, and market demand data.

[0013] Furthermore, in step S3, the inventory correlation analysis model is constructed based on a multiple linear regression algorithm. Data from the entire supply chain data resource pool is used as input variables, and five dependent variables are set: inventory replenishment cycle, inventory consumption rate, inventory transfer timeliness, commodity demand structure, and near-expiry commodity grading results. Through model training, the correlation between supply chain data and inventory replenishment cycle, sales data and inventory consumption rate, logistics data and inventory transfer timeliness, user data and commodity demand structure, and commodity attribute data and near-expiry commodity grading results are analyzed. The model parameters are iteratively optimized using a gradient descent algorithm to determine the influence weights of each independent variable on the dependent variable, forming standardized correlation analysis rules.

[0014] Furthermore, in step S3, the inventory correlation analysis model is constructed based on a multiple linear regression algorithm, and its model formula is as follows: ,in, The inventory replenishment cycle refers to the total time from triggering a replenishment order to the completion of the goods being received into the warehouse. It is the inventory depletion rate, which refers to the amount of inventory of a certain commodity that decreases per unit of time; It refers to the time from when the transfer order is issued to when the goods are delivered to the target store and put into storage. It refers to the demand structure of goods, which is the proportion of demand for different categories / specifications of goods within a certain period; This is the classification result of near-expiry goods, which is quantified through numerical mapping: 0 represents the value preservation disposal level, 1 represents the promotional clearance level, and 2 represents the donation and redemption level. For supplier capacity data, For raw material supply cycle data, This is data related to supply batch size restrictions; For real-time sales data of stores, This refers to the remaining inventory data for goods. This is data on the intensity of promotional activities (quantified as a range of 0-1, where 0 represents no promotion and 1 represents the most intense promotion). For warehouse sorting timeliness data, For cross-regional transportation cycle data, For logistics capacity data; This data represents user consumption preferences (quantified as a range of 0-1, with higher values ​​indicating a stronger preference for that type of product). For repurchase frequency data, Data on regional demand distribution; This refers to the remaining shelf life of the product. This is historical loss rate data. This is market demand data (quantified as continuous values ​​in the range of 0-10). It is the regression coefficient, representing the first... The class dependent variable corresponding to the first The weights of the influence of each independent variable on the dependent variable , , It is the regression constant term, used to correct for model prediction bias. .

[0015] Furthermore, in step S4, the preset inventory safety threshold includes the minimum inventory threshold and the maximum inventory threshold of the product, and the near-expiration classification threshold includes the first threshold and the second threshold related to the remaining shelf life, the preset loss threshold and the warning loss threshold related to the historical loss rate, and the preset demand threshold and the minimum demand threshold related to market demand.

[0016] Furthermore, in step S4, the initial inventory optimization plan includes a product replenishment plan, an inventory transfer plan, a slow-moving product handling plan, and a near-expiry product tiered disposal plan: the product replenishment plan determines the replenishment time and replenishment batch based on supply chain data and inventory consumption rate; the inventory transfer plan determines the transfer direction, transfer quantity, and transportation route based on differences in inventory balances across different stores, logistics data, and regional demand distribution data; the slow-moving product handling plan determines the discount level and promotional channels based on the duration of product inventory backlog and market demand data; and the near-expiry product tiered disposal plan classifies products into grades according to tiering rules and formulates corresponding strategies.

[0017] Furthermore, in step S4, the near-expiry product grading and disposal scheme classifies products into grades according to grading judgment rules and formulates corresponding strategies. The judgment rules are as follows: when the remaining shelf life is greater than the first threshold, the historical loss rate is lower than the preset loss threshold, and the market demand is higher than the preset demand threshold, it is classified as a value preservation disposal grade, and the disposal method of adjusting the display position and bundling sales is adopted; when the remaining shelf life is between the first threshold and the second threshold, and the historical loss rate is between the preset loss threshold and the warning loss threshold, it is classified as a promotional clearance grade, and the discount rate setting method based on dynamic pricing logic is adopted; when the remaining shelf life is less than the second threshold, the historical loss rate is higher than the warning loss threshold, or the market demand is lower than the minimum demand threshold, it is classified as a donation and verification grade, and the automatic docking process of public welfare channels is adopted.

[0018] On the other hand, a retail inventory optimization system based on end-to-end data includes:

[0019] The data acquisition module is used to build a full-link data acquisition channel. It collects supply chain data, sales data, logistics data, user data, and product attribute data through RESTful interfaces, TCP / IP protocol data acquisition interfaces, WebService interfaces, compliant crawler interfaces, and JDBC interfaces, respectively.

[0020] The data preprocessing module is used to perform hierarchical preprocessing on the collected multi-dimensional data from the entire data chain, including outlier removal, missing value supplementation, format standardization conversion, and classification and storage operations, to build a full-chain data resource pool.

[0021] The correlation analysis module is used to build an inventory correlation analysis model based on the multiple linear regression algorithm. It inputs data from the full-chain data resource pool into the model, analyzes the correlation between data of each dimension and inventory status and near-expiry product classification, determines the influence weight of data of each dimension, and forms standardized correlation analysis rules.

[0022] The scheme generation module is used to generate an initial inventory optimization scheme based on the output of the inventory correlation analysis model, combined with the preset inventory safety threshold and near-expiry classification threshold. This scheme includes a product replenishment scheme, an inventory transfer scheme, a slow-moving product handling scheme, and a near-expiry product classification disposal scheme.

[0023] The dynamic closed-loop module is used to collect dynamic data of retail business and execution data of near-expiry goods disposal in real time. The dynamic data is input into the inventory correlation analysis model to adjust the initial inventory optimization plan, and the execution data of near-expiry goods disposal is fed back to the upstream supply chain. Finally, the final inventory optimization plan is sent to the corresponding execution terminal through the message queue.

[0024] Compared with existing technologies, this retail inventory optimization method and system based on end-to-end data has the following advantages:

[0025] This invention establishes a full-link data acquisition channel and utilizes multiple interfaces to collaboratively collect multi-dimensional data from the entire retail business process. This enables systematic management of multi-dimensional retail business data, providing a comprehensive and accurate data foundation for inventory management. The collected data undergoes layered preprocessing to effectively remove abnormal data, supplement missing data, and store it in a unified format. A full-link data resource pool is constructed, ensuring data quality and storage efficiency. The constructed inventory correlation analysis model, based on a multiple linear regression algorithm, clarifies the correlation logic between multi-dimensional data, inventory status, and near-expiry product grading, determines the influence weight of each variable, and forms standardized correlation analysis rules, making inventory decisions more scientific and rational. Furthermore, the near-expiry product grading and disposal system classifies products into different levels according to preset rules and formulates corresponding strategies, establishing a standardized near-expiry product handling process, improving the efficiency and effectiveness of near-expiry product disposal. The supply chain closed-loop feedback mechanism collects dynamic data and near-expiry product disposal execution data in real time, dynamically adjusts the initial inventory optimization plan, and feeds the disposal data back to the upstream supply chain, achieving collaborative linkage between inventory management and upstream procurement, thus improving the technical system of retail inventory management.

[0026] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0028] Figure 1 This is a flowchart of a retail inventory optimization method based on end-to-end data.

[0029] Figure 2 This is a flowchart illustrating the steps involved in constructing and training a correlation analysis model for retail inventory optimization based on end-to-end data.

[0030] Figure 3 This is a structural diagram of a retail inventory optimization system based on end-to-end data. Detailed Implementation

[0031] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0032] This invention provides a retail inventory optimization method and system based on end-to-end data. It establishes an end-to-end data acquisition channel to collaboratively collect multi-dimensional data from the entire retail business process, and performs layered preprocessing on the data to construct an end-to-end data resource pool. Simultaneously, it constructs an inventory correlation analysis model to clarify the correlation logic between multi-dimensional data and inventory status and the classification of near-expiry goods. Based on the model output, it generates an initial inventory optimization plan that includes solutions for replenishment, inventory allocation, handling of slow-moving goods, and graded disposal of near-expiry goods. Furthermore, it establishes a dynamic adjustment and supply chain closed-loop feedback mechanism to adjust the plan in real time and feed the disposal data back to the upstream supply chain, effectively solving problems such as insufficient data integration, lack of standardization in the disposal of near-expiry goods, and lack of supply chain collaboration in traditional technologies.

[0033] This invention provides a retail inventory optimization method based on end-to-end data, such as... Figure 1 As shown below, the steps are explained in detail in the context of a multi-store, multi-category retail scenario:

[0034] S1: Establish a full-chain data collection channel to collaboratively collect data from the supply chain, sales, logistics, user, and product attribute categories according to preset rules;

[0035] Supply chain data collection: Connect to the Supplier Management System (SMS) via a RESTful interface, preset the collection period to off-peak hours to avoid system congestion, and pull supplier capacity data, raw material supply cycle data, and supply batch restriction data;

[0036] Sales data collection: Based on the TCP / IP protocol, build a data collection interface for store sales terminals (POS machines), set a fixed data upload frequency, and collect real-time sales data, merchandise inventory data, and promotional activity data of each store in real time. The intensity of promotional activities is quantified to a specified range according to rules.

[0037] Logistics data collection: Through the WebService interface, we connect with the logistics management system (WMS) to obtain logistics data in real time, including warehouse sorting timeliness data, cross-regional transportation cycle data, and logistics capacity data;

[0038] User-side data collection: Collect relevant data of registered users on the platform through compliant web crawler interfaces, and integrate it with the membership system data to obtain user consumption preference data, repurchase frequency data, and regional demand distribution data. User consumption preferences are quantified to a specified range according to rules.

[0039] Product attribute data collection: Through the JDBC interface, the product information management system (PMS) is connected to extract product attribute data in batches at fixed periods, including product shelf life data, historical loss rate data, and market demand data. Market demand is quantified to a specified range according to rules.

[0040] S2: Perform hierarchical preprocessing on the collected multi-dimensional data across the entire data chain to build a full-chain data resource pool based on a master-slave architecture distributed database;

[0041] Layered classification: Based on data source, the data is divided into supply chain layer, sales layer, logistics layer, user layer, and product attribute layer. Each layer of data is further subdivided into real-time data, near real-time data, and historical data. Different types of data are set with storage periods according to preset rules, and historical data is archived and stored.

[0042] Outlier removal: The 3σ principle is used to remove outlier data. Outlier data that exceeds the normal fluctuation range and deviates from the mean by a preset multiple is automatically marked and removed.

[0043] Missing value imputation: Numerical data is imputed using the mean, and categorical data is imputed using the mode, to ensure data integrity.

[0044] Standardized data format conversion: Unify data format, convert time format according to preset standard specifications, retain decimal places for numerical data with uniform precision, and uniformly map relevant quantitative data to the corresponding preset range.

[0045] Categorized storage: The master database stores real-time / near real-time data, supporting high-concurrency read and write operations, while the slave database stores historical data for model training. This read-write separation improves data access efficiency and builds a full-link data resource pool.

[0046] S3: Construct an inventory correlation analysis model based on a multiple linear regression algorithm, such as... Figure 2 As shown, the association weights of variables are determined through model training, and standardized association analysis rules are formed.

[0047] Model input and output definitions:

[0048] Input variables: taken from the end-to-end data resource pool, including supplier capacity, raw material supply cycle, and supply batch limits on the supply chain side; real-time sales data of stores, remaining inventory, and intensity of promotional activities on the sales side; warehousing and sorting timeliness, cross-regional transportation cycle, and logistics capacity on the logistics side; user consumption preferences, repurchase frequency, and regional demand distribution on the user side; and product attributes such as remaining shelf life, historical loss rate, and market demand.

[0049] Output dependent variables: inventory replenishment cycle, inventory consumption rate, inventory transfer timeliness, commodity demand structure, and near-expiry commodity classification results;

[0050] Model Formula and Parameter Initialization: The model formula uses a system of multiple linear regression equations, and the regression coefficients and regression constants are initialized with preset default values; the model formula is as follows: ,in, The inventory replenishment cycle refers to the total time from triggering a replenishment order to the completion of the goods being received into the warehouse. It is the inventory depletion rate, which refers to the amount of inventory of a certain commodity that decreases per unit of time; It refers to the time from when the transfer order is issued to when the goods are delivered to the target store and put into storage. It refers to the demand structure of goods, which is the proportion of demand for different categories / specifications of goods within a certain period; This is the classification result of near-expiry goods, which is quantified through numerical mapping: 0 represents the value preservation disposal level, 1 represents the promotional clearance level, and 2 represents the donation and redemption level. For supplier capacity data, For raw material supply cycle data, This is data related to supply batch size restrictions; For real-time sales data of stores, This refers to the remaining inventory data for goods. The intensity of the promotional activity is quantified into a range of 0-1, where 0 represents no promotion and 1 represents the most intense promotion. For warehouse sorting timeliness data, For cross-regional transportation cycle data, For logistics capacity data; User consumption preference data is quantified into a 0-1 range, with higher values ​​indicating a stronger preference for that type of product. For repurchase frequency data, Data on regional demand distribution; This refers to the remaining shelf life of the product. This is historical loss rate data. Market demand data is quantified as continuous values ​​in the range of 0-10. It is the regression coefficient, representing the first... The class dependent variable corresponding to the first The weights of the influence of each independent variable on the dependent variable , , It is the regression constant term, used to correct for model prediction bias.

[0051] Model training process:

[0052] Training data: Historical data from a certain period of time are selected and divided into training set and validation set according to a preset ratio;

[0053] Optimization algorithm: The gradient descent algorithm is used to iteratively optimize the model parameters, with the goal of minimizing the mean square error between the model's predicted values ​​and the actual values;

[0054] Weight determination: Through iterative training, the influence weight of each variable on the corresponding dependent variable is determined;

[0055] Standardized association analysis rule generation: Based on the trained weights, standardized rules are formulated to associate changes in each variable with changes in the corresponding dependent variable.

[0056] S4: Based on the output of the inventory correlation analysis model, combined with the preset inventory safety threshold and near-expiry classification threshold, an initial inventory optimization plan is generated;

[0057] Preset threshold setting:

[0058] Inventory safety thresholds: The minimum inventory threshold is set according to a preset ratio based on the average daily consumption rate, and the maximum inventory threshold is set according to another preset ratio based on the average daily consumption rate.

[0059] Near-expiry grading thresholds: The first threshold is set based on the product's shelf life according to a preset ratio, and the second threshold is set based on the product's shelf life according to another preset ratio; at the same time, preset loss thresholds, warning loss thresholds, preset demand thresholds, and minimum demand thresholds are also set.

[0060] Product replenishment plan: The replenishment parameters are determined based on the inventory replenishment cycle and inventory consumption rate. Replenishment is triggered when the current inventory level is lower than the replenishment trigger threshold. The replenishment batch and replenishment time are determined in combination with the supply batch limit.

[0061] Inventory transfer plan: Developed based on differences in store inventory balances, inventory transfer timeliness, and regional demand distribution, determining the transfer direction, transfer quantity, and optimized transportation route;

[0062] Solution for handling slow-moving goods: Based on the length of time the inventory has been held up and the level of market demand, when the length of time the inventory has been held up exceeds the preset slow-moving judgment period and the level of market demand is lower than the minimum demand threshold, the discount level and promotion channels will be determined.

[0063] Grading and Disposal Plan for Near-Expiry Goods:

[0064] Value Preservation Level: When the remaining shelf life of a product is greater than the first threshold, the historical loss rate is lower than the preset loss threshold, and the market demand is higher than the preset demand threshold, the following measures will be taken: adjusting the display location and bundling sales.

[0065] Promotional clearance level: When the remaining shelf life of the product is between the first threshold and the second threshold, and the historical loss rate is between the preset loss threshold and the warning loss threshold, a discount rate setting method based on dynamic pricing logic is adopted.

[0066] Donation verification level: When the remaining shelf life of the goods is less than the second threshold, the historical loss rate is higher than the warning loss threshold, or the market demand is lower than the minimum demand threshold, the automatic connection process through public welfare channels is adopted.

[0067] S5: Real-time collection of dynamic data on retail operations and execution data on the disposal of near-expiry goods, enabling dynamic adjustment of initial inventory optimization plans, closed-loop feedback of the supply chain, and final distribution of the final plan to the end user.

[0068] Dynamic data collection: Real-time collection of dynamic data on retail operations and data on the disposal of near-expiry goods, collected at a fixed frequency.

[0069] Solution Adjustment: Input dynamic data into the inventory correlation analysis model, recalculate the dependent variable, and adjust the initial inventory optimization plan to ensure that the plan is adapted to the real-time business scenario.

[0070] Supply chain closed-loop feedback: Data on the disposal of near-expiry goods is fed back to the upstream supply chain, providing a reference for suppliers to adjust supply batches and production dates.

[0071] Terminal delivery: The final optimization plan is delivered to the corresponding execution terminals via message queues, including the procurement system, logistics scheduling system, store sales terminals, and online mall backend.

[0072] like Figure 3 As shown, this system is built based on the above method and deployed using a distributed architecture. The specific implementation of each module is as follows:

[0073] Data acquisition module: Deployed on a cloud server, it integrates a RESTful interface plugin, TCP / IP data acquisition service, WebService client, compliant crawler framework and JDBC driver, and supports real-time concurrent data acquisition from multiple stores and multiple suppliers.

[0074] Data preprocessing module: Based on a distributed computing framework, it uses distributed computing to perform layered data processing. Operations such as outlier removal and missing value filling are implemented through a data processing library. The processed data stream is stored in a master-slave distributed database to ensure data synchronization between the master and slave databases.

[0075] Association analysis module: Implements a multiple linear regression algorithm based on a data analysis library. Gradient descent optimization is implemented through custom code. The model training task is deployed on a dedicated computing server to improve training efficiency and control the model prediction error within a preset range.

[0076] Solution generation module: Integrates a rule engine, solidifies preset thresholds and correlation analysis rules into a rule library, automatically generates initial optimization solutions through engine reasoning, controls the solution generation time within a preset time, and supports manual intervention and adjustment.

[0077] Dynamic closed-loop module: Deploys real-time data monitoring services to monitor dynamic data of retail business and the progress of solution execution; realizes asynchronous transmission of supply chain feedback data through message queues, and finally distributes the solution to each execution terminal through network interface to ensure the success rate of distribution.

[0078] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A retail inventory optimization method based on end-to-end data, characterized in that, The method includes the following specific steps: S1: Establish a full-chain data collection channel to collaboratively collect data from the supply chain, sales, logistics, user, and product attribute categories according to preset rules; S2: Perform hierarchical preprocessing on the collected multi-dimensional data across the entire data chain to build a full-chain data resource pool based on a master-slave architecture distributed database; S3: Construct an inventory correlation analysis model based on the multiple linear regression algorithm, determine the correlation weights of variables through model training, and form standardized correlation analysis rules; S4: Based on the output of the inventory correlation analysis model, combined with the preset inventory safety threshold and near-expiry classification threshold, an initial inventory optimization plan is generated; S5: Real-time collection of dynamic data on retail operations and execution data on the disposal of near-expiry goods, enabling dynamic adjustment of initial inventory optimization plans, closed-loop feedback of the supply chain, and final distribution of the final plan to the end user.

2. The retail inventory optimization method based on end-to-end data according to claim 1, characterized in that, In step S1, the end-to-end data acquisition channel includes five types of interfaces: a RESTful interface that interfaces with the supplier management system to pull data from the supply chain at a preset period; a data acquisition interface for store sales terminals to collect sales data based on the TCP / IP protocol; a WebService interface that interfaces with the logistics management system to obtain logistics data in real time; and a crawler interface for user data to collect relevant user data. The JDBC interface connects to the product information management system to extract product attribute data in batches.

3. The retail inventory optimization method based on end-to-end data according to claim 1, characterized in that, In step S1, the multi-dimensional data includes supply chain data, sales data, logistics data, user data, and product attribute data. Supply chain data includes supplier capacity data, raw material supply cycle data, and supply batch restriction data. Sales data includes real-time sales data of each store, product inventory balance data, and promotional activity data. Logistics data includes warehousing and sorting timeliness data, cross-regional transportation cycle data, and logistics capacity data. User data includes user consumption preference data, repurchase frequency data, and regional demand distribution data. Product attribute data includes product shelf life data, historical loss rate data, and market demand data.

4. The retail inventory optimization method based on end-to-end data according to claim 1, characterized in that, In step S3, the inventory correlation analysis model is constructed based on a multiple linear regression algorithm. Data from the end-to-end data resource pool is used as input variables, and five dependent variables are set: inventory replenishment cycle, inventory consumption rate, inventory transfer timeliness, commodity demand structure, and near-expiry commodity grading results. The model is trained to analyze the correlation between supply chain data and inventory replenishment cycle, sales data and inventory consumption rate, logistics data and inventory transfer timeliness, user data and commodity demand structure, and commodity attribute data and near-expiry commodity grading results. The model parameters are iteratively optimized using a gradient descent algorithm to determine the influence weight of each independent variable on the dependent variable, thus forming standardized correlation analysis rules.

5. The retail inventory optimization method based on end-to-end data according to claim 1, characterized in that, In step S3, the inventory correlation analysis model is constructed based on the multiple linear regression algorithm, and its model formula is as follows: ,in, The inventory replenishment cycle refers to the total time from triggering a replenishment order to the completion of the goods being received into the warehouse. It is the inventory depletion rate, which refers to the amount of inventory of a certain commodity that decreases per unit of time; It refers to the time from when the transfer order is issued to when the goods are delivered to the target store and put into storage. It refers to the demand structure of goods, which is the proportion of demand for different categories / specifications of goods within a certain period; This is the classification result of near-expiry goods, which is quantified through numerical mapping: 0 represents the value preservation disposal level, 1 represents the promotional clearance level, and 2 represents the donation and redemption level. For supplier capacity data, For raw material supply cycle data, This is data related to supply batch size restrictions; For real-time sales data of stores, This refers to the remaining inventory data for goods. Data on the intensity of promotional activities; For warehouse sorting timeliness data, For cross-regional transportation cycle data, For logistics capacity data; For user consumption preference data, For repurchase frequency data, Data on regional demand distribution; This refers to the remaining shelf life of the product. This is historical loss rate data. For market demand data, It is the regression coefficient, representing the first... The class dependent variable corresponding to the first The weights of the influence of each independent variable on the dependent variable , , It is the regression constant term, used to correct for model prediction bias. .

6. The retail inventory optimization method based on end-to-end data according to claim 1, characterized in that, In step S4, the preset inventory safety threshold includes the minimum inventory threshold and the maximum inventory threshold for the product, and the near-expiration grading threshold includes the first threshold and the second threshold related to the remaining shelf life, the preset loss threshold and the warning loss threshold related to the historical loss rate, and the preset demand threshold and the minimum demand threshold related to market demand.

7. The retail inventory optimization method based on end-to-end data according to claim 1, characterized in that, In step S4, the initial inventory optimization plan includes a product replenishment plan, an inventory transfer plan, a slow-moving product handling plan, and a near-expiry product tiered disposal plan: the product replenishment plan determines the replenishment time and replenishment batch based on supply chain data and inventory consumption rate; the inventory transfer plan determines the transfer direction, transfer quantity, and transportation route based on differences in inventory balances in different stores, logistics data, and regional demand distribution data; the slow-moving product handling plan determines the discount level and promotional channels based on the length of product inventory backlog and market demand data; and the near-expiry product tiered disposal plan classifies products into grades according to tiering rules and formulates corresponding strategies.

8. The retail inventory optimization method based on end-to-end data according to claim 1, characterized in that, In step S4, the near-expiry product grading and disposal scheme classifies products into grades according to grading judgment rules and formulates corresponding strategies. The judgment rules are as follows: when the remaining shelf life is greater than the first threshold, the historical loss rate is lower than the preset loss threshold, and the market demand is higher than the preset demand threshold, it is classified as the value preservation disposal level, and the disposal method of adjusting the display position and bundling sales is adopted; when the remaining shelf life is between the first threshold and the second threshold, and the historical loss rate is between the preset loss threshold and the warning loss threshold, it is classified as the promotion clearance level, and the discount rate setting method based on dynamic pricing logic is adopted; when the remaining shelf life is less than the second threshold, the historical loss rate is higher than the warning loss threshold, or the market demand is lower than the minimum demand threshold, it is classified as the donation and verification level, and the automatic connection process of public welfare channels is adopted.

9. A retail inventory optimization system based on end-to-end data, the system being applicable to the retail inventory optimization method based on end-to-end data as described in any one of claims 1-8, characterized in that, The system includes: The data acquisition module is used to build a full-link data acquisition channel. It collects supply chain data, sales data, logistics data, user data, and product attribute data through RESTful interfaces, TCP / IP protocol data acquisition interfaces, WebService interfaces, compliant crawler interfaces, and JDBC interfaces, respectively. The data preprocessing module is used to perform hierarchical preprocessing on the collected multi-dimensional data from the entire data chain, including outlier removal, missing value supplementation, format standardization conversion, and classification and storage operations, to build a full-chain data resource pool. The correlation analysis module is used to build an inventory correlation analysis model based on the multiple linear regression algorithm. It inputs data from the full-chain data resource pool into the model, analyzes the correlation between data of each dimension and inventory status and near-expiry product classification, determines the influence weight of data of each dimension, and forms standardized correlation analysis rules. The scheme generation module is used to generate an initial inventory optimization scheme based on the output of the inventory correlation analysis model, combined with the preset inventory safety threshold and near-expiry classification threshold. This scheme includes a product replenishment scheme, an inventory transfer scheme, a slow-moving product handling scheme, and a near-expiry product classification disposal scheme. The dynamic closed-loop module is used to collect dynamic data of retail business and execution data of near-expiry goods disposal in real time. The dynamic data is input into the inventory correlation analysis model to adjust the initial inventory optimization plan, and the execution data of near-expiry goods disposal is fed back to the upstream supply chain. Finally, the final inventory optimization plan is sent to the corresponding execution terminal through the message queue.

Images