Intelligent supply chain optimization management method and system based on digital AI analysis

Abstract

The invention relates to the technical field of data processing, and discloses an intelligent supply chain optimization management method and system based on digital AI analysis. The method comprises the steps of collecting supply chain data to construct a resource library; analyzing a sales mode and K-means + + clustering customer features by using LSTM, and predicting demands; inventory parameters and supplier scores are calculated, and a collaborative data set is formed; commodity association is analyzed through a knowledge graph, and a commodity selection decision is generated; resource allocation is carried out by applying multi-objective optimization, and cross-channel inventory collaboration is realized; and an integral rule and a merchant enabling scheme are constructed based on the transaction data, and a digital asset closed-loop mechanism is formed. According to the intelligent supply chain optimization management method based on AI analysis, global visualization, dynamic scheduling and accurate allocation of inventory resources are realized, the technical problem of cross-channel inventory collaborative allocation is solved, and the operation efficiency and the resource utilization rate of the whole supply chain are remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a smart supply chain optimization management method and system based on digital AI analysis. Background Art

[0002] With the rapid development of the digital economy and the widespread application of information technology, traditional supply chain management models are undergoing profound changes. Existing supply chain management technologies primarily focus on inventory management, demand forecasting, logistics distribution, and resource planning, using information systems to collect and process data. Early supply chain management relied primarily on ERP systems for information integration, traditional statistical methods for demand forecasting, the classic EOQ (economic order quantity) model for inventory control, and logistics management systems for distribution planning. With the rise of big data technologies, some leading companies have begun applying methods such as data mining and machine learning to demand forecasting and customer analysis, improving forecast accuracy. Furthermore, the development of the Internet of Things (IoT) enables real-time tracking of the flow of goods within the supply chain, while cloud computing provides technical support for large-scale data processing. These technologies have significantly improved supply chain transparency and responsiveness.

[0003] However, existing technologies still have many shortcomings. First, the various links in traditional supply chain management are often independent information islands, lacking full-link data integration and collaboration, resulting in poor information flow and difficulty in achieving synergy between links. Second, existing demand forecasting methods are not adaptable enough to diverse and personalized market demands, especially when dealing with nonlinear and highly volatile market changes. Third, traditional inventory management models generally adopt a single-channel management approach, which cannot effectively address inventory coordination issues in an omni-channel sales environment, resulting in low overall inventory utilization. Fourth, existing resource scheduling methods lack effective handling of multi-objective optimization, making it difficult to strike a balance between cost, efficiency, and service quality. Finally, existing technologies do not fully explore the value of data assets and are unable to convert the massive data generated in the supply chain into direct commercial value, and the potential of data assets has not been fully released. These problems seriously restrict the improvement of supply chain management efficiency and the enhancement of corporate competitiveness. Summary of the Invention

[0004] This application provides an intelligent supply chain optimization management method and system based on digital AI analysis, which is used to achieve global visualization, dynamic scheduling and precise allocation of inventory resources by establishing an intelligent supply chain optimization management method based on AI analysis, solve the technical difficulties of cross-channel inventory collaborative allocation, and significantly improve the overall supply chain operation efficiency and resource utilization.

[0005] In the first aspect, the present application provides a smart supply chain optimization management method based on digital AI analysis: collecting real-time data from each link of the supply chain, establishing a distributed database cluster for storage, and generating a structured supply chain data resource library; using historical sales data and customer transaction data in the supply chain data resource library, analyzing the time series sales pattern and K-means++ clustering to process customer behavior characteristics through the LSTM algorithm, generating market demand trend forecast results and accurate customer portrait data; based on the market demand trend forecast results and the accurate customer portrait data, calculating the optimal inventory level parameters and determining the supplier scoring weights to generate an intelligent supply chain collaborative data set; inputting the inventory status records and supplier scoring results in the intelligent supply chain collaborative data set into the product ontology analysis process, extracting the product association relationship through the knowledge graph algorithm, and obtaining an intelligent product selection decision parameter table; performing dynamic resource planning calculations on the product combination list generated under the guidance of the intelligent product selection decision parameter table, and applying a multi-objective optimization method to generate fulfillment resource allocation instructions to achieve cross-channel inventory collaborative allocation; based on the transaction data and customer feedback information generated in the cross-channel inventory collaborative allocation process, performing data asset value quantification calculations, and constructing a digital asset closed-loop operation mechanism for point allocation rules and merchant empowerment solutions.

[0006] In a second aspect, the present application provides a smart supply chain optimization and management system based on digital AI analysis, which includes:

[0007] The collection module is used to collect real-time data from all links of the supply chain, establish a distributed database cluster for storage, and generate a structured supply chain data resource library;

[0008] A generation module is used to utilize the historical sales data and customer transaction data in the supply chain data resource library to analyze the time series sales pattern through the LSTM algorithm and process the customer behavior characteristics through K-means++ clustering to generate market demand trend forecast results and accurate customer portrait data;

[0009] a calculation module, configured to calculate optimal inventory level parameters and determine supplier scoring weights based on the market demand trend forecast results and the precise customer profile data, thereby generating an intelligent supply chain collaboration dataset;

[0010] An input module is used to input the inventory status records and supplier rating results in the intelligent supply chain collaborative dataset into the product ontology analysis process, extract product association relationships through the knowledge graph algorithm, and obtain an intelligent product selection decision parameter table;

[0011] An allocation module is used to perform dynamic resource planning calculations on the product combination list generated under the guidance of the intelligent product selection decision parameter table, and to generate fulfillment resource allocation instructions using a multi-objective optimization method to achieve cross-channel inventory collaborative allocation;

[0012] The feedback module is used to perform quantitative calculations on the value of data assets based on the transaction data and customer feedback information generated during the cross-channel inventory collaborative allocation process, and to build a digital asset closed-loop operation mechanism for point allocation rules and merchant empowerment solutions.

[0013] On the third aspect, a smart supply chain optimization management device based on digital AI analysis is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the smart supply chain optimization management device based on digital AI analysis executes the above-mentioned smart supply chain optimization management method based on digital AI analysis.

[0014] In a fourth aspect, a computer-readable storage medium is provided, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the above-mentioned intelligent supply chain optimization management method based on digital AI analysis.

[0015] In the technical solution provided by this application, by collecting real-time data from all links of the supply chain and establishing a distributed database cluster, a structured supply chain data resource library is generated, which lays a solid foundation for full-link data analysis, solves the problem of information islands in traditional supply chains, and significantly improves the integrity and consistency of data; by using historical sales data and customer transaction data in the supply chain data resource library, the time series sales pattern is analyzed through the LSTM algorithm, which gives full play to the advantages of the LSTM algorithm in processing long-term dependencies and effectively captures the seasonality, periodicity and trend characteristics of sales data. At the same time, the application of the K-means++ clustering algorithm in the processing of customer behavior characteristics improves the defect of traditional K-means that is easily affected by the initial center point, improves the clustering stability, and thus generates more accurate market demand trend forecast results and customer portrait data; based on these high-quality forecast results and portrait data, by calculating the optimal inventory level parameters and determining the supplier evaluation By weighting, an intelligent supply chain collaborative data set is generated, which effectively solves the problems of blind procurement and improper supplier selection in traditional inventory management; the intelligent supply chain collaborative data set is combined with product ontology analysis and knowledge graph algorithm to extract product association relationships and obtain an intelligent product selection decision parameter table. The application of knowledge graph algorithm realizes in-depth mining of implicit associations between products, far exceeding the analysis depth of traditional association rule mining; dynamic resource planning is carried out on the product combination list generated under the guidance of the intelligent product selection decision parameter table, and multi-objective optimization method is applied to generate fulfillment resource allocation instructions, which successfully solves the limitations of single-objective optimization in traditional resource scheduling and achieves a balance between cost control, service quality and resource efficiency; finally, based on the transaction data and customer feedback information generated in the process of cross-channel inventory collaborative allocation, data asset value quantitative calculation is performed, and points allocation rules and merchant empowerment solutions are constructed, forming a complete digital asset closed-loop operation mechanism and maximizing data value. This solution fully considers the characteristics of various AI algorithms and their application value in various links of the supply chain, such as the long-term memory capability of the LSTM algorithm in time series forecasting, the optimized initialization characteristics of K-means++ in clustering, the semantic reasoning ability of the knowledge graph in expressing complex relationships, and the equilibrium ability of the multi-objective optimization algorithm in processing multi-dimensional decision spaces. The organic combination of these algorithmic features has enabled this solution to achieve significant technical results in cross-channel inventory collaboration, accurate market forecasting, intelligent resource scheduling, and data value realization. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 This is a schematic diagram of an embodiment of a smart supply chain optimization management method based on digital AI analysis in an embodiment of the present application;

[0018] Figure 2 This is a schematic diagram of an embodiment of an intelligent supply chain optimization management system based on digital AI analysis in an embodiment of the present application;

[0019] Figure 3 It is a schematic block diagram of the structure of an intelligent supply chain optimization management device based on digital AI analysis in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The embodiments of the present application provide a method and system for optimizing and managing an intelligent supply chain based on digital AI analysis. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0021] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiments of the present application, an embodiment of the intelligent supply chain optimization management method based on digital AI analysis includes:

[0022] Step S101: Collect real-time data from all links of the supply chain, establish a distributed database cluster for storage, and generate a structured supply chain data resource library;

[0023] Step S102: Utilize historical sales data and customer transaction data in the supply chain data repository, analyze time series sales patterns through the LSTM algorithm, and process customer behavior characteristics through K-means++ clustering to generate market demand trend forecast results and accurate customer profile data;

[0024] Step S103: Calculate optimal inventory level parameters and determine supplier rating weights based on market demand trend forecast results and accurate customer profile data to generate an intelligent supply chain collaboration dataset;

[0025] Step S104: Input the inventory status records and supplier rating results in the intelligent supply chain collaborative dataset into the product ontology analysis process, extract the product association relationships through the knowledge graph algorithm, and obtain the intelligent product selection decision parameter table;

[0026] Step S105: Perform dynamic resource planning calculations on the product combination list generated under the guidance of the intelligent product selection decision parameter table, apply a multi-objective optimization method to generate fulfillment resource allocation instructions, and achieve cross-channel inventory collaborative allocation;

[0027] Step S106: Based on the transaction data and customer feedback information generated during the cross-channel inventory collaborative allocation process, perform quantitative calculations on the value of data assets, and build a digital asset closed-loop operation mechanism for point allocation rules and merchant empowerment solutions.

[0028] It is understandable that the execution subject of this application can be a smart supply chain optimization management system based on digital AI analysis, or a terminal or server, which is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0029] Specifically, real-time supply chain data is collected through customer flow sensors, product tag identifiers, smart checkout counters, and environmental monitoring devices. Customer flow sensors record customer entry times and stays, product tag identifiers track product flow, smart checkout counters record transaction data, and environmental monitoring devices collect in-store temperature and humidity data. This raw data is cleaned to remove outliers, fill in missing values, and remove duplicate records to create a standardized data structure. It is then categorized into hot and cold data, with hot data stored on high-speed devices and cold data migrated to low-cost storage. An incremental data synchronization mechanism is established to ensure data consistency. Standard APIs are used to connect with enterprise ERP, CRM, and WMS systems, ultimately generating a structured supply chain data repository. Analysis is conducted using this repository. First, historical sales data is arranged into time series and external factors such as holiday information and weather data are incorporated to construct a sales time series dataset. This dataset is divided into training, validation, and test sets. A sliding window is used to construct sequence pairs, which are then input into an LSTM network. The LSTM network consists of an embedding layer, multiple hidden layers, and an output layer. An attention mechanism is employed to assess the importance of different time points and influencing factors. The network optimizes weight parameters using a gradient descent algorithm to generate market demand trend forecasts. Customer transaction data is then subjected to Z-score normalization and principal component analysis for dimensionality reduction, forming an eigenvector matrix. This matrix is ​​then clustered using the K-means++ algorithm, using the elbow rule to determine the optimal number of clusters. Clustering calculations are then performed to generate customer group classifications. This result is then combined with consumption time, frequency, and amount to calculate a customer value score, generating precise customer profile data.

[0030] Based on the market demand trend forecast results, volatility was analyzed. A multi-objective optimization model was established based on product attributes. A genetic algorithm was used to calculate the economic order quantity and safety stock, resulting in a preliminary inventory parameter plan. This plan was adjusted based on the product seasonality index and a probabilistic inventory model was used to address inventory risks for long-tail products, generating optimal inventory level parameters. Furthermore, indicators such as on-time delivery rate and product qualification rate were extracted from historical supplier collaboration data. The analytic hierarchy process was used to construct an evaluation index system, determining the scoring weights for each indicator. The scoring weights were multiplied by the fuzzy evaluation matrix to generate the supplier scoring results. Based on the optimal inventory level parameters and supplier scoring results, a procurement decision matrix was constructed. A linear programming algorithm was used to optimize supplier selection and order allocation plans, generating an intelligent collaborative procurement plan. This plan was then matched and analyzed with precise customer profile data to form an intelligent supply chain collaboration dataset.

[0031] The inventory status records and supplier rating results from the intelligent supply chain collaboration dataset are processed. First, a unified product information entry specification is established, standardizing attributes such as category and brand to generate a basic product information table. This table is then combined with the supplier rating results to construct a product ontology model. Semantic association techniques are used to define hierarchical relationships between products, forming an initial product knowledge graph. Inference rules are applied to this graph, identifying implicit associations between products through transitivity calculations and path analysis, generating a product association network. Based on this network and historical sales data, an association rule mining algorithm is used to calculate support, confidence, and lift metrics between products, generating a product portfolio recommendation index. This index is then cross-matched with precise customer profile data to assess the degree of product-customer segment fit and generate a product marketability score matrix. This matrix is ​​then combined with product lifecycle and gross profit margin data for a multi-dimensional evaluation. A decision tree algorithm is then used to generate a product classification strategy, resulting in a table of intelligent product selection decision parameters.

[0032] Based on the intelligent product selection decision parameter table, a product combination list is generated. Order volume is predicted based on historical order data, and a resource demand matrix is ​​constructed to generate an initial resource demand plan. Capacity estimates are calculated for warehousing, manpower, and transportation resources in this plan, load thresholds are set, and baseline resource allocation values ​​are calculated to form a basic resource planning scheme. This scheme is then input into a multi-objective optimization model, with target functions for cost, service timeliness, and resource utilization. Resource allocation combinations are generated using a Pareto optimal set, resulting in a dynamic resource planning solution. Based on this solution, an order allocation decision tree is constructed to assess order urgency, delivery distance, and resource availability, generating fulfillment resource allocation instructions. Based on this instruction, inventory across all channels is globally mapped to establish an inventory sharing pool. Cross-channel orders are processed using an inventory allocation rules engine to generate a cross-channel inventory transfer plan. This plan implements virtual inventory consolidation and physical separation controls, maintaining data consistency through real-time synchronization, enabling cross-channel inventory collaborative allocation. Transaction data and customer feedback generated during cross-channel inventory collaborative allocation are categorized and organized, and a data asset catalog is created, annotating information such as type and source. Based on this list, a value assessment method is applied, scoring data assets based on completeness, accuracy, and other dimensions to generate a data asset value score. Based on this score and consumer transaction behavior, a consumer points allocation model is constructed, with weights assigned to consumption amount and frequency to derive a consumer points allocation rule. This rule is applied to the transaction system, and a points acquisition and redemption process is designed. This process is recorded on the blockchain, and a points account management system is established. Merchants' digitalization capabilities are assessed based on system application capabilities and data management capabilities, generating a digital profile of the merchant. Based on this profile and the points account management system, a merchant empowerment resource package is constructed, allocating differentiated digital tools and service resources to different merchant types, forming a closed-loop operation mechanism for digital assets.

[0033] In the embodiment of the present application, by collecting real-time data from all links of the supply chain and establishing a distributed database cluster, a structured supply chain data resource library is generated, which lays a solid foundation for full-link data analysis, solves the problem of information islands in traditional supply chains, and significantly improves the integrity and consistency of data; by using historical sales data and customer transaction data in the supply chain data resource library, the time series sales pattern is analyzed through the LSTM algorithm, which gives full play to the advantages of the LSTM algorithm in processing long-term dependencies and effectively captures the seasonality, cyclicality and trend characteristics of sales data. At the same time, the application of the K-means++ clustering algorithm in the processing of customer behavior characteristics improves the defect of traditional K-means that is easily affected by the initial center point, improves the clustering stability, and thus generates more accurate market demand trend forecast results and customer portrait data; based on these high-quality forecast results and portrait data, by calculating the optimal inventory level parameters and determining the supplier scoring weights The intelligent supply chain collaborative dataset is generated, which effectively solves the problems of blind procurement and improper supplier selection in traditional inventory management; the intelligent supply chain collaborative dataset is combined with product ontology analysis and knowledge graph algorithm to extract product association relationships and obtain an intelligent product selection decision parameter table. The application of knowledge graph algorithm realizes the deep mining of implicit associations between products, which far exceeds the analysis depth of traditional association rule mining; dynamic resource planning is carried out on the product combination list generated under the guidance of the intelligent product selection decision parameter table, and the multi-objective optimization method is applied to generate fulfillment resource allocation instructions, which successfully solves the limitations of single-objective optimization in traditional resource scheduling and achieves a balance between cost control, service quality and resource efficiency; finally, based on the transaction data and customer feedback information generated in the process of cross-channel inventory collaborative allocation, the data asset value is quantitatively calculated, and the points allocation rules and merchant empowerment solutions are constructed, forming a complete digital asset closed-loop operation mechanism and maximizing the data value. This solution fully considers the characteristics of various AI algorithms and their application value in various links of the supply chain, such as the long-term memory capability of the LSTM algorithm in time series forecasting, the optimized initialization characteristics of K-means++ in clustering, the semantic reasoning ability of the knowledge graph in expressing complex relationships, and the equilibrium ability of the multi-objective optimization algorithm in processing multi-dimensional decision spaces. The organic combination of these algorithmic features has enabled this solution to achieve significant technical results in cross-channel inventory collaboration, accurate market forecasting, intelligent resource scheduling, and data value realization.

[0034] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0035] Original supply chain data is obtained by collecting store transaction data through customer flow sensors, product tag identifiers, and smart checkout counters, and recording in-store environmental parameters using environmental monitoring equipment.

[0036] Perform data cleaning on the original supply chain data, remove outliers, fill in missing values, and remove duplicate records to generate a standardized supply chain data structure;

[0037] Categorize standardized supply chain data structures into hot and cold data, allocate high-speed storage device resources to hot data, and migrate cold data to low-cost storage units to form a tiered storage architecture.

[0038] Establish an incremental data synchronization mechanism based on a tiered storage architecture, set up data encryption transmission channels and access control rules, and build a supply chain data security protection system;

[0039] Execute feature extraction algorithms on data within the supply chain data security protection system, extract key attributes and feature vectors from the data, and output a supply chain feature data matrix;

[0040] Connect the supply chain feature data matrix with the enterprise ERP, CRM, and WMS management tools through a standardized API interface to generate a structured supply chain data resource library.

[0041] Specifically, customer flow sensors collect data on customer movement and dwell time within stores. These sensors are typically installed at store entrances, main aisles, and popular areas, recording customer location information every few seconds. Product tag readers, using RFID or QR code technology, record product flow data from storage to sales, including product codes, location change times, and status changes. Smart checkout counters collect transaction-related data, including details such as product type, quantity, price, payment method, and membership information. Environmental monitoring equipment records in-store environmental parameters such as temperature, humidity, and light levels, which are potentially correlated with customer behavior and product sales. These four types of equipment operate simultaneously, forming a raw supply chain data stream. Cleaning the collected raw supply chain data is a key step in ensuring data quality. Outlier processing involves identifying and removing data outside a reasonable range, such as records with negative or unusually large transaction amounts and sensor data that clearly does not conform to reality. Missing values ​​are imputed using different strategies depending on the data type. For time series data, previous value filling or linear interpolation are used, while for categorical data, mode filling is used. Duplicate records are removed by comparing the data's timestamps and key identifiers, retaining the latest or most complete record and deleting the remaining copies. After these cleaning steps, the raw data is transformed into a standardized supply chain data structure with a consistent format and reliable content.

[0042] Categorizing standardized supply chain data structures into hot and cold data is an important means of optimizing storage resource utilization. Hot data refers to frequently accessed, recent data, such as sales records from the past week, current inventory status, and real-time customer flow data. This data is allocated to high-speed storage devices such as solid-state drives or in-memory databases to ensure fast read and write responses. Cold data refers to historical data with less frequent access, such as transaction records from the past few months or even years. This type of data is migrated to low-cost storage units such as mechanical hard drive arrays or cloud storage to reduce storage costs. Data classification is evaluated based on access frequency, data creation time, and business importance, forming a tiered storage architecture that strikes a balance between performance and cost.

[0043] An incremental data synchronization mechanism is established based on a tiered storage architecture, transmitting and updating only the changed data, rather than the entire data set, reducing network transmission burdens. During data transmission, an encrypted transmission channel is established using the TLS / SSL protocol to encrypt data and prevent theft or tampering during transmission. Access control rules are also implemented to assign appropriate data access rights based on user roles. For example, sales personnel can only access sales-related data, while warehouse managers can only access inventory data, ensuring data security. These security measures collectively constitute a supply chain data security system, safeguarding data security during storage and transmission.

[0044] Feature extraction algorithms are executed on data within the supply chain data security system to extract key information from massive amounts of data. Specifically, sales data is extracted for time features (sales period, periodicity), product features (category, price range), and customer features (consumption level, loyalty); inventory data is extracted for features such as turnover rate, out-of-stock rate, and slow-selling rate; and supplier data is extracted for features such as on-time delivery rate, product qualification rate, and response speed. Principal component analysis is used during feature extraction to reduce data dimensionality, retain the most representative features, and reduce redundant information. The extracted data forms a supply chain feature data matrix, which contains the most valuable information from the original data and facilitates subsequent analysis and application. The supply chain feature data matrix is ​​connected to the company's existing information management systems via standardized APIs to enable data sharing and business collaboration. This includes integration with ERP systems to share inventory, procurement, and sales data; integration with CRM systems to share customer information and consumer behavior data; and integration with WMS systems to share warehousing and logistics data. The standardized API interface adopts a RESTful design style to ensure consistent data exchange standards and compatible formats across different systems. Through this docking and integration, a structured supply chain data resource library is formed, providing a comprehensive and accurate data foundation for subsequent AI analysis and intelligent decision-making.

[0045] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0046] Arrange and clean the historical sales data in the supply chain data repository by time series, and build a sales time series dataset by combining holiday information, weather data, and industry indexes;

[0047] The sales time series dataset is divided into training set, validation set and test set. Sequence pairs are constructed through sliding windows and input into the LSTM network structure to obtain the time series sales pattern analysis results.

[0048] An attention mechanism is introduced into the analysis results of time-series sales patterns, gating parameters and activation functions are set, and weights are optimized through gradient descent to generate market demand trend forecast results.

[0049] Standardize and reduce the dimension of customer transaction data in the supply chain data repository, retain the main feature dimensions, and form a customer feature vector matrix;

[0050] Apply the K-means++ clustering algorithm to the customer feature vector matrix, determine the number of clusters using the elbow rule, set iteration parameters, execute clustering, and generate customer behavior feature classification results;

[0051] Combine the customer behavior characteristics classification results with the consumption time, frequency, and amount dimensions for analysis, assign weight coefficients, calculate the value score, and form accurate customer portrait data.

[0052] Specifically, historical sales data from the supply chain data repository is arranged and cleaned according to time series, sorting the sales data by time granularity, such as year, month, day, and time period, to form a continuous time series. This cleaning process includes processing abnormal sales records, such as sudden surges caused by promotional activities or outliers caused by system errors. Holiday information (such as New Year's Day, Spring Festival, and National Day) is then added to the time series as feature markers, as these special dates are often accompanied by significant changes in sales patterns. Weather data (such as temperature, humidity, and precipitation) is also integrated to identify external factors influencing sales. Macroeconomic indicators such as industry indices such as the Consumer Confidence Index and the Retail Sales Climate Index are also added to reflect the overall market environment. This multidimensional data set forms a sales time series dataset, which contains a complete record of sales volume, sales revenue, and various influencing factors.

[0053] Splitting a constructed sales time series dataset into training, validation, and test sets is a standard step in machine learning model development. This split is typically performed in a ratio of 7:2:1 or 8:1:1 to ensure sufficient training data richness and representative test data. Temporal continuity must be considered during the split to prevent future data from leaking into the training set. Sequence pairs are then constructed using a sliding window approach. The input window length (e.g., the previous 30 days of data) and the prediction window length (e.g., the next 7 days) are set, and a large number of input-output pairs are generated through sliding. These sequence pairs are then fed into a long short-term memory (LSTM) network for training. An LSTM is a specialized recurrent neural network with an input gate, a forget gate, and an output gate structure capable of learning long-term dependencies. The network receives the input sequence, controls the flow of information through an internal gating mechanism, and outputs the corresponding prediction results. After multiple rounds of training, the network gradually learns seasonal patterns, cyclical changes, and long-term trends in the sales data, generating time series sales pattern analysis results.

[0054] Introducing an attention mechanism into the results of time-series sales pattern analysis is a means to further improve forecasting accuracy. This attention mechanism allows the model to dynamically focus on different parts of the input sequence when generating forecasts, assigning higher weights to key time points. For example, for seasonal products, the model will prioritize sales performance over the same period last year; for products affected by holidays, the model will prioritize sales performance during historical holiday periods. Attention weights are determined by calculating the similarity between the query vector and the key vector, typically using a dot product or additive attention method. Setting the gating parameters involves initializing the weight matrices for each gate unit in the LSTM, including the input gate, forget gate, and output gate. Activation functions typically used are the sigmoid function (for gate units) and the tanh function (for state updates and outputs). Model training utilizes a gradient descent algorithm, calculating the gradient of the loss function with respect to each parameter through backpropagation and updating the model weights accordingly. An early stopping strategy is used during training to monitor the error on the validation set to avoid overfitting. The optimized model ultimately generates market demand trend forecasts, including sales volume forecasts for each future time period and confidence intervals for the forecasts.

[0055] Analysis of customer transaction data is fundamental to building accurate customer profiles. Customer transaction data in the supply chain data repository is standardized to bring features of varying dimensions (such as purchase amount, purchase frequency, and membership points) to the same scale. Common methods include Z-score normalization (converting the data to a distribution with a mean of 0 and a standard deviation of 1) or Min-Max normalization (scaling the data to the interval [0, 1]). Normalized data typically has high dimensionality, necessitating dimensionality reduction techniques to reduce the number of features and computational complexity. Principal component analysis (PCA) is a commonly used linear dimensionality reduction method. It projects the original high-dimensional data into a low-dimensional space by finding the direction with the largest variance. Dimensionality reduction requires identifying the primary feature dimensions to retain, typically selecting the top few principal components with a cumulative explained variance of 85% to 95%. After standardization and dimensionality reduction, a customer feature vector matrix is ​​formed, with each row representing a customer and each column representing a processed feature dimension.

[0056] The K-means++ clustering algorithm is used to discover natural clusters within customer groups. Compared to traditional K-means, K-means++ optimizes the selection of initial centroids, maximizing the distance between them. This improves clustering stability and results quality. The algorithm first determines the number of clusters, K, which is typically determined using the elbow rule. The elbow rule calculates the within-cluster sum of squares (WCSS) for different K values, plots the WCSS versus K curve, and identifies the "elbow" point where the curve clearly bends, identifying the optimal K value. Iterative parameters include the maximum number of iterations (typically 100 to 300) and a convergence threshold (e.g., 1e-4). Iterations terminate when the centroid distance falls below the threshold or the maximum number of iterations is reached. During the clustering process, the algorithm continuously adjusts the centroid positions and assigns each customer to the category represented by the nearest centroid. Ultimately, customer behavior characteristics are classified, with each customer assigned to a specific group with similar consumer behavior characteristics.

[0057] Combining the results of customer behavior feature classification with specific consumer behavior indicators can form a more accurate customer profile. The consumption time dimension includes time characteristics such as the time of first consumption, the time of last consumption, and the interval between consumptions; the consumption frequency dimension includes the number of consumptions, the average interval between consumptions, and the regularity of consumption; and the consumption amount dimension includes the total consumption amount, the average single consumption amount, and the highest consumption amount. Different weight coefficients are assigned to these three dimensions, typically based on business importance and predictive power. For example, R (last consumption time), F (consumption frequency), and M (consumption amount) are assigned weights of 0.3, 0.3, and 0.4, respectively. Customer value scores are calculated through weighted summation, with customers with high scores representing high-value and high-potential customer groups. Based on the comprehensive score and clustering results, accurate customer profile data is formed, including multi-dimensional characteristics such as customer spending power, consumption preferences, price sensitivity, and loyalty.

[0058] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0059] Conduct a volatility analysis on the market demand trend forecast results, establish a multi-objective optimization model based on product attribute data, calculate the economic order quantity and safety stock through genetic algorithms, and obtain a preliminary inventory parameter plan;

[0060] Based on the preliminary inventory parameter plan and the seasonal index of the commodity, an adjustment coefficient is applied to seasonal commodities, and the inventory risk of long-tail commodities is processed through the probabilistic inventory model to generate the optimal inventory level parameters;

[0061] We extracted the supplier's historical cooperation data on on-time delivery rate, product qualification rate, price competitiveness, and response speed, and used the analytic hierarchy process to construct an evaluation index system to obtain the scoring weights of each index.

[0062] Multiply the weight of each indicator by the fuzzy comprehensive evaluation matrix to conduct a risk assessment of the supplier's financial status, production stability, and industry reputation to form a supplier scoring result;

[0063] Based on the optimal inventory level parameters and supplier rating results, a procurement decision matrix is ​​constructed. The supplier selection and order allocation schemes are optimized through a linear programming algorithm to generate an intelligent collaborative procurement plan.

[0064] Match and analyze intelligent collaborative procurement plans with precise customer portrait data, establish a collaborative mechanism for demand forecasting and supply response, and form an intelligent supply chain collaborative data set.

[0065] Specifically, volatility analysis focuses on the magnitude, frequency, and pattern of demand fluctuations. These characteristics are quantified by calculating the standard deviation, coefficient of variation, and autocorrelation coefficient of demand. The standard deviation reflects the degree of dispersion in demand, the coefficient of variation eliminates the impact of differences in sales volume across products, and the autocorrelation coefficient reflects the correlation between demand at different points in time. A multi-objective optimization model is developed by combining product attribute data (such as product category, life cycle stage, profit margin, volume and weight, and shelf life). This model simultaneously considers three conflicting objectives: inventory holding cost, stock-out cost, and ordering cost. The multi-objective optimization problem is solved using a genetic algorithm. Genetic algorithms simulate natural selection and genetic processes, continuously evolving the population through operations such as encoding, selection, crossover, and mutation to search for the optimal solution. The economic order quantity is calculated by taking into account factors such as demand, ordering cost, and holding cost; the safety stock calculation considers factors such as service level, demand volatility, and replenishment lead time. During the iterative calculation process of the genetic algorithm, the quality of each solution is evaluated by the fitness function, high-quality solutions are retained through selection operations, and new solutions are generated through crossover and mutation operations. After multiple generations of evolution, a set of better economic order quantity and safety stock parameters are obtained to form a preliminary inventory parameter plan.

[0066] Further optimization of the initial inventory parameter plan is necessary, especially for products with significant seasonal characteristics. This is combined with a product seasonality index, typically derived by decomposing time series, which reflects seasonal variations in sales intensity. Adjustment coefficients are applied to seasonal products, increasing inventory levels before peak sales seasons and reducing them during off-seasons to align inventory levels with seasonal demand fluctuations. Specifically, the seasonality index is multiplied by the initial inventory parameters to obtain inventory parameters that account for seasonal factors. Long-tail products, which are products with low sales volumes but a wide variety of products, present significant challenges and carry high risk in inventory management. Probabilistic inventory models are used to address inventory risks for long-tail products. These models account for the random nature of demand and calculate the optimal inventory level for a given service level. Based on probability distributions of demand, such as the normal or Poisson distribution, probabilistic inventory models calculate the expected total cost or expected profit at different inventory levels to determine the optimal inventory level. After seasonal adjustment and addressing long-tail products, more accurate optimal inventory level parameters are generated.

[0067] Supplier evaluation is an important part of smart supply chain management. First, the supplier's historical cooperation data is processed to extract key evaluation indicators, including delivery on-time rate (the proportion of actual on-time deliveries to the total number of deliveries), product qualification rate (the proportion of qualified products to the total supply quantity), price competitiveness (comparison of supplier prices with the average market price) and response speed (the average time from order placement to delivery). The evaluation index system is constructed using the hierarchical analysis method. This method decomposes complex problems into multi-level element groups by establishing a hierarchical structure, and then compares them pairwise to determine the relative importance of each indicator. The specific operation steps include: establishing a hierarchical model, arranging the evaluation objectives, evaluation criteria and evaluation objects in layers; constructing a judgment matrix, and comparing the elements of the same layer pairwise using the 1-9 ratio scaling method; calculating the weight vector, and using the eigenvalue method or logarithmic least squares method to solve the eigenvector of the judgment matrix; consistency test to ensure the rationality of the judgment; and finally obtaining the scoring weight of each indicator.

[0068] Multiply the weights of each indicator obtained in the previous step by the fuzzy comprehensive evaluation matrix to conduct a comprehensive supplier evaluation. Fuzzy comprehensive evaluation is an effective method for handling ambiguity and uncertainty in evaluation. It reflects the affiliation between evaluation indicators and evaluation levels by establishing a fuzzy relationship matrix. First, the evaluation indicator set and evaluation level set are determined. Then, based on the actual situation, the degree of membership of each indicator corresponding to each evaluation level is determined to form a fuzzy relationship matrix. A fuzzy combination operation is performed on the indicator weight vector and the fuzzy relationship matrix to obtain a comprehensive evaluation result. In addition to conventional indicators, a risk assessment is conducted on the supplier's financial status, production stability, and industry reputation. These factors are directly related to the stability and continuity of the supply chain. The risk assessment results are combined with the conventional indicator assessment results to form a supplier rating result, providing a quantitative basis for supplier selection. Based on the optimal inventory level parameters and the supplier rating results, a procurement decision matrix is ​​constructed. The rows of this matrix represent different products, the columns represent different suppliers, and the matrix elements represent corresponding procurement decision variables, such as purchase quantity. Supplier selection and order allocation are optimized using a linear programming algorithm. Linear programming is a classic mathematical optimization method used to find the optimal value of a linear objective function under a set of linear constraints. In this scenario, the objective function is typically minimizing total procurement cost or maximizing supply quality, while constraints include demand fulfillment, supplier capacity, and budget constraints. Linear programming algorithms, such as the simplex method, iteratively find the optimal solution, determining which products to purchase from which suppliers and in what quantities, and generating an intelligent collaborative procurement plan.

[0069] Intelligent collaborative procurement plans are matched and analyzed with precise customer profile data to ensure that purchased products meet the needs of the target customer groups. This matching analysis involves aligning the characteristics of purchased products with customer preferences, such as high-end products with high-value customers, and seasonal products with seasonal consumer groups. Through this matching analysis, a coordinated mechanism for demand forecasting and supply response is established to achieve precise matching of supply and demand. This mechanism not only considers historical demand patterns but also the dynamic nature of customer behavior changes, enabling dynamic adjustments to procurement plans based on the latest customer profile data. The result is an intelligent supply chain collaborative dataset that integrates inventory optimization parameters, supplier evaluation results, procurement plans, and customer demand data, providing comprehensive data support for smart supply chain management.

[0070] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0071] Based on the inventory status records in the intelligent supply chain collaborative data set, a unified commodity information entry specification is established, and the category, brand, specification, origin, and material are standardized to obtain a commodity basic information table;

[0072] Combine the basic product information table and supplier rating results to build a product ontology model. Use semantic association technology to define the hierarchical relationship and attribute constraints between products to form an initial product knowledge graph.

[0073] Apply graph reasoning rules to the initial product knowledge graph, identify implicit associations between products through transitive calculations and path analysis algorithms, and generate a product association network;

[0074] Based on the product association network and historical sales data, an association rule mining algorithm is used to calculate the support, confidence, and lift indicators between products to obtain the product combination recommendation index;

[0075] Cross-match the product combination recommendation index with accurate customer portrait data, apply the weighted cosine similarity algorithm to evaluate the matching degree between products and customer groups, and generate a product marketability scoring matrix;

[0076] A multi-dimensional evaluation is conducted on the product marketability scoring matrix combined with the product life cycle stage and gross profit margin data, and a product classification strategy is generated using the decision tree algorithm to obtain an intelligent product selection decision parameter table.

[0077] Specifically, a unified product information entry specification was established based on inventory status records in the intelligent supply chain collaborative dataset. This specification integrated product data from different systems and resolved data inconsistencies. Category data was standardized to establish a multi-level product classification system. For example, "women's jeans" was standardized into a hierarchical structure of "apparel > women's clothing > trousers > jeans." Brand data was standardized, with the spelling and capitalization of brand names standardized to resolve the issue of multiple representations of the same brand. For example, "P&G" and "Procter & Gamble" were unified into a single identifier. Specifications were standardized, with unified units of measurement and representation, such as "500ml" and "0.5L" being standardized. Origin information was standardized, with a unified mapping between region codes and names established. Material information was categorized and organized, and a material type dictionary was created. This standardization process resulted in a structured product basic information table containing fields such as product ID, standardized category, standardized brand, standardized specification, standardized origin, and standardized material.

[0078] The product basic information table and supplier rating results are combined to construct a product ontology model. An ontology model is a formal description of domain concepts and their relationships, used to represent the conceptual system and semantic relationships of products. During the construction process, the product class hierarchy is first defined, clarifying hierarchical relationships. Next, product attributes are defined, including data type attributes (such as price and weight) and object attributes (such as supplier and manufacturer). Attribute constraints, such as value ranges and cardinality limits, are defined. Finally, product instances are created, mapping specific products into the ontology structure. Semantic association techniques are used to define hierarchical relationships and attribute constraints between products. Semantic association techniques leverage the semantic similarity and contextual relationships of words to identify and establish conceptual associations between products. For example, by analyzing product description text, it can be determined that "apple" can refer to both a fruit category and an electronic product brand, determining the correct semantic meaning based on the context. The ontology model constructed in this way forms an initial product knowledge graph. This graph represents product concepts, entities, and their relationships in a graphical structure, with nodes representing products or attributes and edges representing relationships between them. Graph inference rules are applied to the initial product knowledge graph to discover implicit product associations. Graph inference rules include both logical and statistical inference rules. Transitive calculations are applied based on logical reasoning rules. For example, "Product A is a substitute for Product B, and Product B is a substitute for Product C" can be deduced to "Product A is a substitute for Product C." Path analysis algorithms identify specific path patterns in the graph, calculate path similarities, and discover implicit associations. For example, by analyzing the path pattern of "Product A → Usage Scenario → Product B," it was found that although Products A and B belong to different categories, they are applicable to the same scenarios and have a complementary relationship. Through these reasoning methods, the implicit relationships in the knowledge graph are supplemented to generate a more complete product association network, which includes not only direct product associations (such as the same category or brand), but also implicit product associations (such as complementary, substitution, and matching relationships).

[0079] Based on the product association network and historical sales data, association rule mining algorithms are used to calculate the strength of associations between products. Association rule mining is a method for discovering implicit relationships between sets of items. The most commonly used algorithms are the Apriori algorithm and the FP-Growth algorithm. These algorithms scan transaction datasets, identify frequent itemsets, and then generate association rules. Three key metrics are calculated for each pair of products: support (the probability that two items are purchased together), confidence (the probability that item B is also purchased when item A is purchased), and lift (indicating the degree to which purchasing item A influences purchasing item B; a value greater than 1 indicates a positive correlation). For example, by analyzing transaction records in historical sales data to identify patterns in which "bread" and "milk" are frequently purchased together, their support, confidence, and lift are calculated. A high support indicates a common combination, a high confidence indicates a strong association, and a high lift indicates a non-chance association. By combining these three metrics, a product combination recommendation index is derived, which quantifies the strength of associations between products and provides a basis for intelligent product selection.

[0080] The product portfolio recommendation index is cross-matched with precise customer profile data to assess the degree of match between products and customer groups. Precise customer profile data includes customer demographics, consumption behavior, and preferences. A weighted cosine similarity algorithm is used to calculate the similarity between product feature vectors and customer preference vectors. Weighted cosine similarity takes into account the varying importance of different features, assigning higher weights to important features. The calculation process first extracts product features (such as price range, category, and style) and customer preference features (such as historical purchase categories, price sensitivity, and brand preference). Feature vectors are then constructed, and finally, weighted cosine similarity is calculated. Higher similarity values ​​indicate a stronger match between products and customer groups. This matching analysis generates a product marketability score matrix, where rows represent products, columns represent customer groups, and matrix elements represent corresponding marketability scores.

[0081] Further analysis of the product marketability score matrix, combined with product lifecycle stage and gross profit margin data, led to a multi-dimensional assessment. Product lifecycle stages include introduction, growth, maturity, and decline, and products at different stages require different product selection strategies. Gross profit margin reflects a product's profitability and is a key factor in product selection decisions. This multi-dimensional assessment comprehensively considers the marketability score, lifecycle stage, and gross profit margin to generate the final product selection evaluation metric. A decision tree algorithm is used to generate a product classification strategy. A decision tree is a tree-structured classification model that recursively partitions a dataset to form decision rules. In this solution, the decision tree construction process includes selecting the optimal feature as the split point (such as the marketability score); partitioning the dataset based on the feature value; and recursively repeating the above process for each sub-dataset until a stopping condition is reached (such as reaching the maximum depth or insufficient node samples). The resulting decision tree can classify products into categories such as "must-have," "optional," "trial-sale," and "eliminate," forming a table of intelligent product selection decision parameters. This parameter table specifies a clear product selection strategy for each category of goods, including parameters such as purchase quantity, shelf location, and promotion intensity.

[0082] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0083] Generate a product combination list based on the intelligent product selection decision parameter table, combine order history data to predict order volume, and use a mixed integer programming algorithm to construct a resource demand matrix to obtain an initial resource demand plan.

[0084] Calculate the capacity of warehousing, manpower, and transportation resources in the initial resource demand plan, set the maximum load threshold based on the constraints, and use linear programming methods to calculate the resource allocation benchmark value to form a basic resource planning plan;

[0085] Input the basic resource planning scheme into the multi-objective optimization model, set the objective functions of cost minimization, service time optimization and resource utilization maximization, generate multiple resource configuration combinations through the Pareto optimal set, and obtain the dynamic resource planning results;

[0086] Build an order allocation decision tree based on the dynamic resource planning results, prioritize orders by evaluating order urgency, delivery distance, and resource availability, and generate fulfillment resource allocation instructions;

[0087] Based on fulfillment resource allocation instructions, we conduct a global visual mapping of each channel's inventory, establish an inventory sharing pool, and process cross-channel orders through the inventory allocation rule engine to form a cross-channel inventory transfer plan.

[0088] Implement virtual inventory consolidation and physical separation control for cross-channel inventory transfer plans, maintain inventory data consistency through real-time inventory synchronization mechanism and reservation and release algorithm, and achieve cross-channel inventory collaborative allocation.

[0089] Specifically, a product portfolio list is generated based on the intelligent product selection decision parameter table. This list includes products that have been screened and categorized by the algorithm as "mandatory," "optional," "trial sale," and "obsolete." Each product has corresponding parameters such as recommended purchase quantity, inventory level, and promotional strategy. This product portfolio list is combined with historical order data to predict order volume. Historical order data typically includes order records from a certain period (such as six months or a year) and contains information such as order time, product details, order quantity, and delivery region. Time series analysis methods (such as moving average, exponential smoothing, or ARIMA models) are used to predict order volume for the future. The forecasting process takes into account seasonal, trend, and cyclical factors, resulting in order volume forecasts by category, region, and time period. Based on the forecast results, a mixed integer programming algorithm is used to construct a resource demand matrix. Mixed integer programming is an optimization algorithm whose decision variables include both integer variables (such as the number of equipment and personnel) and continuous variables (such as the proportion of working hours). During the construction process, products are categorized according to their logistics attributes (volume, weight, storage conditions, etc.). The storage space, manpower, and transportation requirements for each product category are calculated, forming a multidimensional matrix. Rows represent product categories, columns represent resource types, and matrix elements represent the quantity of each resource required per unit of product. By solving a mixed integer programming model, the predicted order volume is multiplied by the resource demand matrix to calculate the overall resource requirements, resulting in the initial resource demand plan. Capacity estimation for the storage, manpower, and transportation resources in the initial resource demand plan is a key step in ensuring that resource supply meets demand. Warehousing resource capacity estimation includes warehouse space, shelf capacity, and special storage area capacity. Human resource capacity estimation includes the number of staff during normal working hours, the number of staff available for overtime, and the number of temporary workers available. Transportation resource capacity estimation includes the capacity of owned vehicles, the capacity of partner logistics providers, and the capacity available for emergency use. Based on the company's actual situation and industry experience, maximum capacity thresholds are set for each resource, such as the warehouse's safe storage limit, the maximum workload of staff, and the maximum load capacity of vehicles. Demand exceeding these thresholds is supplemented by external resources or demand smoothing. Linear programming is used to calculate baseline resource allocation values. Linear programming is a method for optimizing a linear objective function under linear constraints. The objective function is typically set to minimize total resource cost, with constraints such as meeting demand, not exceeding resource limits, and adhering to operational rules. By solving the linear programming model, baseline allocation values ​​for various resources are determined, including inventory allocation ratios for each warehouse, staffing requirements for each shift, and vehicle allocations for each delivery route, forming a foundational resource planning plan.

[0090] The basic resource planning solution is input into a multi-objective optimization model to further optimize resource allocation. Multi-objective optimization is an optimization method that simultaneously considers multiple objective functions and is suitable for situations where there are conflicting objectives. In this solution, three objective functions are set: cost minimization (including the sum of warehousing costs, labor costs, and logistics costs), service time optimization (the average time from order receipt to delivery), and resource utilization maximization (the ratio of actual resource usage to total capacity of each resource). These three objectives are often mutually constrained. For example, improving service time often requires increasing resource investment, leading to higher costs, while pursuing high resource utilization may lead to reduced service time. Multi-objective optimization applies the Pareto optimality principle, which states that if a configuration solution cannot improve other objectives without compromising other objectives, it is considered a Pareto optimal solution. Using heuristic algorithms such as genetic algorithms and particle swarm optimization, a series of Pareto optimal solutions are generated, forming multiple resource allocation combinations. Based on the company's strategic priorities and current operating conditions, the most suitable solution is selected from these combinations to produce a dynamic resource planning result. Constructing an order allocation decision tree based on the dynamic resource planning results is the core of intelligent order processing. A decision tree is a tree-structured classification model, where internal nodes represent attribute tests, branches represent test outputs, and leaf nodes represent decision outcomes. During the order allocation decision process, the order's urgency is first assessed. This is done by calculating the difference between the promised delivery time and the current time, and then factoring in the order's priority (such as VIP customer orders, promotional orders, etc.) to determine the urgency score. Next, delivery distance is assessed. The shortest delivery route and required time are determined by calculating the distance from the order's delivery address to each warehouse. Resource availability is then assessed, examining the inventory status of each warehouse, the manpower allocation, and the transportation capacity status of each region to determine the resource conditions necessary to execute the order. Based on these evaluation results, a decision tree algorithm (such as ID3, C4.5, or CART) is used to construct an order allocation decision model. Based on the order's characteristics, this model automatically determines which warehouse to ship from, which logistics method to assign, and when to schedule delivery, generating detailed fulfillment resource allocation instructions.

[0091] Based on fulfillment resource allocation instructions, a global visual mapping of inventory across channels is performed to establish a cross-channel inventory coordination mechanism. Global visual mapping unifies the inventory status of different channels (such as offline stores, regional warehouses, central warehouses, and e-commerce warehouses) to form a complete inventory distribution map. The mapping process requires standardized definitions of inventory status, such as "available inventory," "in-transit inventory," "reserved inventory," and "damaged inventory," to ensure cross-channel data consistency. Based on the global inventory mapping, an inventory sharing pool is established—a collection of inventory resources shared by multiple channels. Inventory within the shared pool is dynamically allocated to each channel based on pre-set rules, improving overall inventory utilization. Cross-channel orders are processed using an inventory allocation rules engine. The rules engine is a rule-based logical judgment tool that automatically makes decisions based on pre-set business rules. Inventory allocation rules include nearest shipment rules, inventory balancing rules, and dedicated warehouse rules for special products. For cross-channel orders, the rules engine determines the optimal inventory transfer path based on order characteristics and current inventory status, creating a cross-channel inventory transfer plan.

[0092] Implement detailed controls over cross-channel inventory transfer plans to achieve virtual and physical inventory coordination. Virtual inventory consolidation manages inventory across different channels as a single entity at the data level, while physical separation maintains independent physical inventory storage. This mechanism enables unified global inventory allocation while maintaining physical inventory decentralization (facilitating rapid response to local demand). Implementing this requires establishing a real-time inventory synchronization mechanism to ensure real-time updates of inventory data across all channels, including sales shipments, purchases, and transfer changes. Furthermore, a reservation-release algorithm is introduced to automatically release unused, but reserved, inventory back to a usable state, avoiding wasted inventory resources. Maintaining inventory data consistency is fundamental to the entire process. Distributed transaction processing, data validation mechanisms, and exception handling procedures ensure synchronized and accurate inventory data across systems. Through these detailed control mechanisms, cross-channel inventory coordination is ultimately achieved, improving overall supply chain responsiveness and resource utilization efficiency.

[0093] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0094] Classify and organize transaction data and customer feedback generated during cross-channel inventory collaborative allocation, establish a data asset catalog, and annotate data types, sources, formats, and ownership relationships to form a data asset list;

[0095] Based on the data asset list, the data value assessment is applied to quantify the score, and the scores of each dimension are summarized through a weighted algorithm to obtain the data asset value score;

[0096] Based on the data asset value score and consumer transaction behavior, a consumption points allocation model is constructed, weight coefficients for consumption amount, frequency, and contribution are set, and the points calculation logic is executed to obtain the consumption points allocation rules;

[0097] Apply consumption points allocation rules to the transaction system, design a process for points acquisition channels and redemption methods, use blockchain technology to record the points issuance and redemption process, and establish a points account management system;

[0098] Conduct a multi-dimensional assessment of merchants' digitalization level, evaluating them from the perspectives of system application capabilities, data management levels, and operational capabilities. Based on the scoring results, merchants are graded and classified to generate digital profiles of merchants.

[0099] Based on the merchant digital portrait files and points account management system, a merchant empowerment resource package is constructed, and differentiated digital tools and service resources are allocated to different types of merchants through intelligent matching algorithms, forming a closed-loop operation mechanism for digital assets.

[0100] Specifically, the transaction data and customer feedback information generated during the cross-channel inventory collaborative allocation process are classified and organized. Transaction data includes order information, payment records, delivery status, and return and exchange records; customer feedback information includes evaluation content, complaint records, consultation content, and usage suggestions. The classification and organization process uses data classification standards to divide data into structured data (such as order amount, payment method) and unstructured data (such as evaluation text, voice feedback). On this basis, a data asset catalog is established, and a unique identifier is assigned to each type of data. The data type (such as transaction type, behavior type, feedback type), data source (such as online shopping mall, physical store, service hotline), data format (such as text, number, image) and ownership relationship (such as corporate ownership, user authorization, third-party permission) are annotated. These annotated information forms the metadata layer, which together with the original data constitutes a complete data asset list, providing a basis for subsequent data value assessment.

[0101] Applying a data value assessment method to quantitatively score data based on a data asset inventory is a core step in data asset management. Data value assessment evaluates data across multiple dimensions, including data completeness (the fill rate and coverage of data fields), data accuracy (the degree to which the data conforms to reality), data timeliness (data update frequency and time sensitivity), data scarcity (data acquisition difficulty and market uniqueness), data application value (the degree to which the data supports business decisions), and data compliance (the legal compliance of data acquisition and use). Scoring criteria are set for each dimension. For example, completeness can be scored by calculating the proportion of non-null fields, while accuracy can be scored by sampling verification to determine the degree of consistency with reality. The raw scores for each dimension are normalized and converted to a unified scale. The scores are then aggregated using a weighted algorithm, with weighting coefficients determined based on business importance and data application scenarios. For example, accuracy and application value are weighted more heavily for data used for customer profiling, while timeliness is weighted more heavily for data used for real-time decision-making. A weighted summation is used to calculate a comprehensive score for each data category, forming a data asset value score table that provides a quantitative basis for data asset management and application.

[0102] Building a consumption points allocation model based on data asset value scores and consumer transaction behavior is a key means of monetizing data value and incentivizing users. This model is based on three core metrics: consumption amount, consumption frequency, and contribution. The consumption amount metric directly reflects a consumer's purchasing power and economic value, typically proportional to the transaction amount. The consumption frequency metric reflects consumer activity and loyalty, measured by the number of purchases per unit time. The contribution metric measures value creation beyond direct consumption, such as providing effective feedback, participating in product reviews, and recommending new users. Weighting coefficients are assigned to these three metrics. Typically, consumption amount has a higher weight, such as 50% of the total score, while consumption frequency and contribution account for 30% and 20%, respectively. Based on the actual values ​​of each metric and the weighting coefficients, the points calculation logic is executed to calculate the appropriate points value for each transaction. This calculation process also takes into account business rules, such as doubling points on holidays, doubling points for specific products, and point multipliers based on membership tiers. These calculation rules ultimately form a comprehensive consumption points allocation system, clearly defining the points reward standards for different consumer behaviors.

[0103] Applying consumption point allocation rules to a transaction system requires detailed process design for point earning and redemption methods. Points can be earned through direct consumption, event participation, social sharing, and surveys; redemption methods include product redemption, service redemption, coupon redemption, and partner point swaps. To ensure the fairness and transparency of the points system, blockchain technology is introduced to record the points issuance and redemption process. Blockchain, a distributed ledger technology with immutable and fully traceable characteristics, is ideally suited for recording points transactions. Specifically, each point issuance or redemption is recorded as a transaction record on the blockchain, containing information such as transaction time, user ID, transaction type, point amount, and transaction description. Blockchain nodes are maintained jointly within the enterprise and by trusted third parties to ensure data security and reliability. Based on blockchain records, a points account management system is established, including functions such as point balance inquiry, point expiration management, point usage restrictions, and point exception handling, providing users with a transparent and convenient points management experience.

[0104] A multi-dimensional assessment of merchants' digital capabilities is a prerequisite for merchant empowerment. This assessment focuses on three key areas: system application capabilities, data management capabilities, and operational capabilities. System application capabilities assess merchants' level of proficiency and use of information systems, including infrastructure configuration (e.g., cash register equipment, inventory management systems), software application capabilities (e.g., ERP usage, mobile application integration), and technical support capabilities (e.g., IT staffing and maintenance response speed). Data management capabilities assess merchants' ability to collect, store, and utilize data, including comprehensiveness of data collection (e.g., customer information collection rate, transaction data integrity), data analysis capabilities (e.g., report generation capabilities, data mining applications), and data security management (e.g., data backup mechanisms, privacy protection measures). Operational capabilities assess merchants' actual business execution efficiency, including sales capabilities (e.g., single-store sales, average order value), inventory management capabilities (e.g., inventory turnover rate, out-of-stock rate), and customer service capabilities (e.g., customer satisfaction, repeat purchase rate). Through this comprehensive assessment, each merchant is assigned a digital score, which is then categorized into four levels: digitally advanced, digitally growing, digitally initiated, and digitally lagging. Generate a detailed digital portrait file for each merchant, including scores in each dimension, analysis of strengths and weaknesses, and suggestions for improvement, providing a basis for subsequent precise empowerment.

[0105] Based on the merchant's digital profile and points account management system, a merchant empowerment resource package is constructed to provide differentiated support for different merchant types. This empowerment resource package includes various resource types, including technical tools (such as SaaS systems and data analysis tools), training resources (such as operation guides and training courses), marketing support (such as traffic allocation and campaign plans), and financial services (such as supply chain finance and credit lines). Differentiated resources are allocated to different merchant types through an intelligent matching algorithm, which calculates matching based on merchant profile characteristics and resource characteristics. Specifically, the algorithm first extracts key feature vectors from the merchant profile, including digitalization level, business scale, and industry category. It then extracts feature vectors for each resource, including applicable conditions, usage difficulty, and expected results. Finally, it calculates the similarity or compatibility between feature vectors to determine the most suitable resource combination for each merchant. Through this precise matching, leading merchants are provided with advanced analytical tools and innovative technology applications, growing merchants are provided with system integration and data application training, entry-level merchants are provided with basic digital tools and operational guidance, and lagging merchants are provided with simple applications and basic training. Intelligent matching and precise empowerment have formed a closed-loop operation mechanism for digital assets, realizing the maximum utilization of data value and the continuous improvement of merchant capabilities.

[0106] The above describes the intelligent supply chain optimization management method based on digital AI analysis in the embodiment of the present application. The following describes the intelligent supply chain optimization management system based on digital AI analysis in the embodiment of the present application. Figure 2 In the embodiments of the present application, an embodiment of a smart supply chain optimization management system based on digital AI analysis includes:

[0107] The collection module 201 is used to collect real-time data from all links of the supply chain, establish a distributed database cluster for storage, and generate a structured supply chain data resource library;

[0108] A generation module 202 is configured to utilize the historical sales data and customer transaction data in the supply chain data repository to analyze time series sales patterns using an LSTM algorithm and process customer behavior characteristics using K-means++ clustering to generate market demand trend forecasts and accurate customer profile data;

[0109] A calculation module 203 is configured to calculate optimal inventory level parameters and determine supplier scoring weights based on the market demand trend forecast results and the precise customer profile data, thereby generating an intelligent supply chain collaboration dataset;

[0110] Input module 204 is used to input the inventory status records and supplier rating results in the intelligent supply chain collaborative dataset into the product ontology analysis process, extract product association relationships through the knowledge graph algorithm, and obtain an intelligent product selection decision parameter table;

[0111] Allocation module 205 is used to perform dynamic resource planning calculations on the product combination list generated under the guidance of the intelligent product selection decision parameter table, and generate fulfillment resource allocation instructions using a multi-objective optimization method to achieve cross-channel inventory collaborative allocation;

[0112] Feedback module 206 is used to perform data asset value quantification calculation based on the transaction data and customer feedback information generated during the cross-channel inventory collaborative allocation process, and to build a digital asset closed-loop operation mechanism for point allocation rules and merchant empowerment solutions.

[0113] Through the collaborative cooperation of the above components, by collecting real-time data from all links of the supply chain and establishing a distributed database cluster, a structured supply chain data resource library is generated, which lays a solid foundation for full-link data analysis, solves the problem of information islands in the traditional supply chain, and significantly improves the integrity and consistency of the data; using the historical sales data and customer transaction data in the supply chain data resource library, the time series sales pattern is analyzed through the LSTM algorithm, which gives full play to the advantages of the LSTM algorithm in dealing with long-term dependencies and effectively captures the seasonality, cyclicality and trend characteristics of sales data. At the same time, the application of the K-means++ clustering algorithm in the processing of customer behavior characteristics improves the defect of traditional K-means that is easily affected by the initial center point, improves the clustering stability, and thus generates more accurate market demand trend forecast results and customer portrait data; based on these high-quality forecast results and portrait data, by calculating the optimal inventory level parameters and determining the supply The weights of merchant scores are calculated to generate an intelligent supply chain collaborative data set, which effectively solves the problems of blind procurement and improper supplier selection in traditional inventory management; the intelligent supply chain collaborative data set is combined with product ontology analysis and knowledge graph algorithm to extract product associations and obtain an intelligent product selection decision parameter table. The application of knowledge graph algorithm realizes the deep mining of implicit associations between products, which far exceeds the analysis depth of traditional association rule mining; dynamic resource planning is carried out on the product combination list generated under the guidance of the intelligent product selection decision parameter table, and multi-objective optimization method is applied to generate fulfillment resource allocation instructions, which successfully solves the limitations of single-objective optimization in traditional resource scheduling and achieves a balance between cost control, service quality and resource efficiency; finally, based on the transaction data and customer feedback information generated in the process of cross-channel inventory collaborative allocation, data asset value quantitative calculation is performed, and points allocation rules and merchant empowerment solutions are constructed, forming a complete digital asset closed-loop operation mechanism and maximizing data value. This solution fully considers the characteristics of various AI algorithms and their application value in various links of the supply chain, such as the long-term memory capability of the LSTM algorithm in time series forecasting, the optimized initialization characteristics of K-means++ in clustering, the semantic reasoning ability of the knowledge graph in expressing complex relationships, and the equilibrium ability of the multi-objective optimization algorithm in processing multi-dimensional decision spaces. The organic combination of these algorithmic features has enabled this solution to achieve significant technical results in cross-channel inventory collaboration, accurate market forecasting, intelligent resource scheduling, and data value realization.

[0114] above Figure 2 The intelligent supply chain optimization management system based on digital AI analysis in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The intelligent supply chain optimization management device based on digital AI analysis in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0115] Figure 3 This is a structural diagram of a smart supply chain optimization management device based on digital AI analysis provided by an embodiment of the present invention. The smart supply chain optimization management device 300 based on digital AI analysis may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more massive storage devices) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or persistent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the smart supply chain optimization management device 300 based on digital AI analysis. Furthermore, the processor 310 can be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the smart supply chain optimization management device 300 based on digital AI analysis to implement the steps of the above-mentioned smart supply chain optimization management method based on digital AI analysis.

[0116] The intelligent supply chain optimization management device 300 based on digital AI analysis may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. It will be understood by those skilled in the art that Figure 3 The structure of the intelligent supply chain optimization management device based on digital AI analysis shown does not constitute a limitation of the intelligent supply chain optimization management device based on digital AI analysis provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0117] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of the intelligent supply chain optimization management method based on digital AI analysis.

[0118] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0119] If the integrated unit is implemented in the form of 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 the present invention is essentially 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. The computer software product is stored in a storage medium and includes several instructions for enabling a smart supply chain optimization management device based on digital AI analysis (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk and other media that can store program code.

[0120] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A smart supply chain optimization management method based on digital AI analysis, characterized by: The method comprises: Collect real-time data from all links of the supply chain, establish a distributed database cluster for storage, and generate a structured supply chain data resource library; Utilizing historical sales data and customer transaction data in the supply chain data repository, the LSTM algorithm is used to analyze time-series sales patterns and K-means++ clustering is used to process customer behavior characteristics, generating market demand trend forecasts and accurate customer profile data. Calculate optimal inventory level parameters and determine supplier scoring weights based on the market demand trend forecast results and the precise customer profile data to generate an intelligent supply chain collaboration dataset; The inventory status records and supplier rating results in the intelligent supply chain collaborative dataset are input into the product ontology analysis process, and the product association relationship is extracted through the knowledge graph algorithm to obtain the intelligent product selection decision parameter table; Perform dynamic resource planning calculations on the product combination list generated under the guidance of the intelligent product selection decision parameter table, apply multi-objective optimization methods to generate fulfillment resource allocation instructions, and achieve cross-channel inventory collaborative allocation; Based on the transaction data and customer feedback information generated in the cross-channel inventory collaborative allocation process, quantitative calculation of data asset value is performed, and a digital asset closed-loop operation mechanism for points allocation rules and merchant empowerment solutions is constructed.

2. The intelligent supply chain optimization management method based on digital AI analysis according to claim 1 is characterized in that: The real-time data of each link in the supply chain is collected, a distributed database cluster is established for storage, and a structured supply chain data resource library is generated, including: Original supply chain data is obtained by collecting store transaction data through customer flow sensors, product tag identifiers, and smart checkout counters, and recording in-store environmental parameters using environmental monitoring equipment. Performing data cleaning on the raw supply chain data to remove outliers, fill in missing values, and remove duplicate records to generate a standardized supply chain data structure; Classify the standardized supply chain data structure into hot data and cold data, allocate high-speed storage device resources to hot data, and migrate cold data to low-cost storage units to form a tiered storage architecture; Establish an incremental data synchronization mechanism based on the hierarchical storage architecture, set up data encryption transmission channels and access control rules, and build a supply chain data security protection system; Execute a feature extraction algorithm on the data within the supply chain data security protection system, extract key attributes and feature vectors from the data, and output a supply chain feature data matrix; The supply chain characteristic data matrix is ​​connected to the enterprise ERP, CRM, and WMS management tools through a standardized API interface to generate a structured supply chain data resource library.

3. The intelligent supply chain optimization management method based on digital AI analysis according to claim 1 is characterized in that: The method utilizes the historical sales data and customer transaction data in the supply chain data repository, analyzes the time series sales pattern through the LSTM algorithm, and processes customer behavior characteristics through K-means++ clustering to generate market demand trend forecast results and accurate customer portrait data, including: Arrange and clean the historical sales data in the supply chain data resource library in time series, and construct a sales time series data set by combining holiday information, weather data, and industry index; The sales time series dataset is divided into a training set, a validation set, and a test set. Sequence pairs are constructed through a sliding window and input into an LSTM network structure to obtain the time series sales pattern analysis results. Introducing an attention mechanism into the analysis results of the time series sales pattern, setting gating parameters and activation functions, optimizing weights through gradient descent, and generating market demand trend forecast results; Standardizing and reducing the dimension of the customer transaction data in the supply chain data repository, retaining the main feature dimensions, and forming a customer feature vector matrix; Applying a K-means++ clustering algorithm to the customer feature vector matrix, determining the number of clusters by the elbow rule, setting iteration parameters, performing clustering processing, and generating customer behavior feature classification results; The customer behavior characteristics classification results are combined with the consumption time, frequency, and amount dimensions for analysis, weight coefficients are assigned, and value scores are calculated to form accurate customer portrait data.

4. The intelligent supply chain optimization management method based on digital AI analysis according to claim 1 is characterized in that: The method of calculating the optimal inventory level parameters and determining the supplier scoring weights based on the market demand trend forecast results and the precise customer profile data to generate an intelligent supply chain collaborative data set includes: Conducting a volatility analysis on the market demand trend forecast results, establishing a multi-objective optimization model based on product attribute data, calculating the economic order quantity and safety stock through a genetic algorithm, and obtaining a preliminary inventory parameter plan; Based on the preliminary inventory parameter plan and the seasonal index of the commodity, an adjustment coefficient is applied to seasonal commodities, and the inventory risk of long-tail commodities is processed through a probabilistic inventory model to generate the optimal inventory level parameters; We extracted the supplier's historical cooperation data on on-time delivery rate, product qualification rate, price competitiveness, and response speed, and used the analytic hierarchy process to construct an evaluation index system to obtain the scoring weights of each index. Multiply the weight of each indicator by the fuzzy comprehensive evaluation matrix to conduct a risk assessment on the supplier's financial status, production stability, and industry reputation to form a supplier scoring result; Based on the optimal inventory level parameters and the supplier rating results, a procurement decision matrix is ​​constructed, supplier selection and order allocation schemes are optimized through a linear programming algorithm, and an intelligent collaborative procurement plan is generated; The intelligent collaborative procurement plan is matched and analyzed with the precise customer portrait data to establish a demand forecasting and supply response collaborative mechanism to form an intelligent supply chain collaborative data set.

5. The intelligent supply chain optimization management method based on digital AI analysis according to claim 1 is characterized in that: The inventory status records and supplier rating results in the intelligent supply chain collaborative dataset are input into the product ontology analysis process, and the product association relationship is extracted through the knowledge graph algorithm to obtain the intelligent product selection decision parameter table, including: Based on the inventory status records in the intelligent supply chain collaborative data set, a unified commodity information entry specification is established, and the category, brand, specification, origin, and material are standardized to obtain a commodity basic information table; Combining the basic product information table with the supplier rating results to construct a product ontology model, defining the hierarchical relationship and attribute constraints between products through semantic association technology, and forming an initial product knowledge graph; Applying graph reasoning rules to the initial product knowledge graph, identifying implicit associations between products through transitive calculations and path analysis algorithms, and generating a product association network; Based on the product association network and historical sales data, an association rule mining algorithm is used to calculate the support, confidence, and lift indicators between products to obtain a product combination recommendation index; Cross-matching the product combination recommendation index with the precise customer portrait data, applying a weighted cosine similarity algorithm to evaluate the degree of matching between products and customer groups, and generating a product marketability scoring matrix; A multi-dimensional evaluation is performed on the product marketability scoring matrix in combination with the product life cycle stage and gross profit margin data, and a product classification strategy is generated using a decision tree algorithm to obtain an intelligent product selection decision parameter table.

6. The intelligent supply chain optimization management method based on digital AI analysis according to claim 1 is characterized in that: The method of performing dynamic resource planning calculations on the product combination list generated under the guidance of the intelligent product selection decision parameter table, and applying a multi-objective optimization method to generate fulfillment resource allocation instructions to achieve cross-channel inventory collaborative allocation includes: Generate a product combination list based on the intelligent product selection decision parameter table, predict order quantity based on historical order data, and construct a resource demand matrix using a mixed integer programming algorithm to obtain an initial resource demand plan; Calculate the capacity of warehousing, manpower, and transportation resources in the initial resource demand plan, set a maximum load threshold based on the constraint conditions, and apply the linear programming method to calculate the resource allocation benchmark value to form a basic resource planning plan; Input the resource planning basic scheme into the multi-objective optimization model, set the objective functions of cost minimization, service time optimization and resource utilization maximization, generate multiple resource configuration combinations through the Pareto optimal set, and obtain the dynamic resource planning results; Building an order allocation decision tree based on the dynamic resource planning results, prioritizing orders by evaluating order urgency, delivery distance, and resource availability, and generating fulfillment resource allocation instructions; Perform a global visual mapping of each channel's inventory based on the fulfillment resource allocation instructions, establish an inventory sharing pool, process cross-channel orders through the inventory allocation rule engine, and form a cross-channel inventory transfer plan; The cross-channel inventory transfer plan is subject to virtual inventory consolidation and physical separation control, and inventory data consistency is maintained through a real-time inventory synchronization mechanism and a reservation and release algorithm to achieve cross-channel inventory collaborative allocation.

7. The intelligent supply chain optimization management method based on digital AI analysis according to claim 1 is characterized in that: The digital asset closed-loop operation mechanism based on transaction data and customer feedback information generated during the cross-channel inventory collaborative allocation process, performing quantitative calculation of data asset value, and constructing points allocation rules and merchant empowerment solutions includes: Classify and organize the transaction data and customer feedback information generated during the cross-channel inventory collaborative allocation process, establish a data asset catalog, annotate the data type, source, format and ownership relationship, and form a data asset list; Based on the data asset list, a data value assessment is applied to perform quantitative scoring, and the scores of each dimension are aggregated through a weighted algorithm to obtain a data asset value score; Based on the data asset value score and consumer transaction behavior, a consumption points allocation model is constructed, weight coefficients of consumption amount, frequency and contribution are set, and point calculation logic is executed to obtain consumption points allocation rules; Apply the consumption points allocation rules to the transaction system, design a process for points acquisition channels and redemption methods, record the points issuance and redemption process through blockchain technology, and establish a points account management system; Conduct a multi-dimensional assessment of merchants' digitalization level, evaluating them from the perspectives of system application capabilities, data management levels, and operational capabilities. Based on the scoring results, merchants are graded and classified to generate digital profiles of merchants. Based on the merchant digital portrait files and the points account management system, a merchant empowerment resource package is constructed, and differentiated digital tools and service resources are allocated to different types of merchants through intelligent matching algorithms, forming a closed-loop operation mechanism for digital assets.

8. A smart supply chain optimization management system based on digital AI analysis, characterized by: A method for implementing a smart supply chain optimization management method based on digital AI analysis according to any one of claims 1 to 7, wherein the smart supply chain optimization management system based on digital AI analysis comprises: The collection module is used to collect real-time data from all links of the supply chain, establish a distributed database cluster for storage, and generate a structured supply chain data resource library; A generation module is used to utilize the historical sales data and customer transaction data in the supply chain data resource library to analyze the time series sales pattern through the LSTM algorithm and process the customer behavior characteristics through K-means++ clustering to generate market demand trend forecast results and accurate customer portrait data; a calculation module, configured to calculate optimal inventory level parameters and determine supplier scoring weights based on the market demand trend forecast results and the precise customer profile data, thereby generating an intelligent supply chain collaboration dataset; An input module is used to input the inventory status records and supplier rating results in the intelligent supply chain collaborative dataset into the product ontology analysis process, extract product association relationships through the knowledge graph algorithm, and obtain an intelligent product selection decision parameter table; An allocation module is used to perform dynamic resource planning calculations on the product combination list generated under the guidance of the intelligent product selection decision parameter table, and to generate fulfillment resource allocation instructions using a multi-objective optimization method to achieve cross-channel inventory collaborative allocation; The feedback module is used to perform quantitative calculations on the value of data assets based on the transaction data and customer feedback information generated during the cross-channel inventory collaborative allocation process, and to build a digital asset closed-loop operation mechanism for point allocation rules and merchant empowerment solutions.

9. A smart supply chain optimization management device based on digital AI analysis, characterized by: It includes a memory and a processor, the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the intelligent supply chain optimization management method based on digital AI analysis as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the processor executes the intelligent supply chain optimization management method based on digital AI analysis as described in any one of claims 1 to 7.

Images