Commodity distribution management method and system based on big data analysis

Abstract

The invention discloses a commodity distribution management method and system based on big data analysis, and relates to the technical field of intelligent distribution management, and the method comprises the steps: collecting multi-dimensional data, and carrying out the preprocessing of the multi-dimensional data; constructing a sales volume prediction model based on a Transform model, inputting the preprocessed multi-dimensional data into the sales volume prediction model, and obtaining the predicted sales volume of the sales terminal; constructing a distribution network graph structure based on the sales volume predicted by the sales terminal; obtaining a distribution scheme through a dynamic graph neural network based on the distribution network graph structure; dynamically adjusting the distribution scheme through multi-objective optimization to generate an optimal distribution scheme; and based on the optimal distribution scheme, customizing a regional distribution scheme through regional adaptive optimization. According to the invention, through integration and preprocessing of multi-dimensional data, sales volume prediction based on a Transform model, dynamic optimization of a distribution network diagram and generation of an optimal distribution scheme of multi-objective optimization, distribution efficiency, flexibility and accuracy are significantly improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent distribution management technology, and in particular to a commodity distribution management method and system based on big data analysis. Background Art

[0002] With the rapid development of big data technology, artificial intelligence and supply chain management technology, the intelligence level of commodity distribution management has been significantly improved. Traditional commodity distribution management methods usually rely on static analysis of historical sales data, inventory information and logistics routes, and make distribution decisions through manually set rules or simple algorithms. This method can meet basic distribution needs when the market environment is relatively stable and the supply and demand relationship is relatively simple. However, with the intensification of market competition, consumer demand has gradually become personalized and diversified. At the same time, driven by the rapid development of e-commerce, the complexity of commodity distribution management has increased significantly. To meet this challenge, existing technologies have gradually introduced big data analysis to provide more accurate decision support for distribution management by integrating multi-dimensional data such as historical sales records, market activities, and distribution costs. However, current distribution management technologies still have certain limitations. For example, existing technologies focus on a single or a few dimensions (such as historical sales and inventory data) in the processing of multi-dimensional data, and have weak integration and analysis capabilities for cross-dimensional data (such as social media sentiment, regional economic changes, and climate factors). In addition, many distribution management systems rely on static rule settings in the optimization of distribution networks, lack dynamic adjustment capabilities, and are difficult to quickly respond to market changes or emergencies (such as logistics disruptions, demand surges, etc.). These problems have greatly limited the efficiency and flexibility of distribution management systems in the current complex and dynamic market environment.

[0003] However, existing technologies have obvious deficiencies in the following two aspects: First, in terms of data integration and feature extraction, existing technologies usually rely only on a single dimension (such as historical sales) or fixed features, and it is difficult to fully capture the multidimensional factors that affect product sales. For example, real-time feedback from consumers on social media, changes in regional economic conditions, seasonal climate fluctuations, etc., may have a significant impact on the sales trend of goods, but traditional technologies lack the ability to effectively integrate and model these data. Second, in terms of optimization and dynamic adjustment of distribution networks, existing technologies are mainly optimized through preset rules or historical experience, and cannot dynamically adjust distribution routes or warehousing scheduling according to real-time changes in market demand, resulting in low distribution efficiency or increased distribution costs. For example, when consumer demand in a certain area suddenly increases, the existing distribution system may not be able to quickly adjust logistics routes or optimize inventory allocation, which in turn leads to an imbalance between supply and demand. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a commodity distribution management method based on big data analysis to solve the problems of insufficient multi-dimensional data integration capabilities and lack of dynamic adjustment capabilities of the distribution network in the prior art.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a commodity distribution management method based on big data analysis, which includes collecting multi-dimensional data and preprocessing the multi-dimensional data; constructing a sales prediction model based on a Transformer model, inputting the preprocessed multi-dimensional data into the sales prediction model, and obtaining the predicted sales at the sales terminal; constructing a distribution network graph structure based on the predicted sales at the sales terminal; based on the distribution network graph structure, obtaining a distribution plan through a dynamic graph neural network; dynamically adjusting the distribution plan through multi-objective optimization to generate an optimal distribution plan; based on the optimal distribution plan, customizing the regional distribution plan through regional adaptive optimization.

[0007] As a preferred solution of the commodity distribution management method based on big data analysis described in the present invention, wherein: the multi-dimensional data includes commodity attribute data, sales terminal data, market environment data, consumer behavior data and distribution cost data; The multi-dimensional data preprocessing includes denoising, normalization, time series processing, feature encoding and feature dimensionality reduction.

[0008] As a preferred solution of the commodity distribution management method based on big data analysis described in the present invention, wherein: based on the Transformer model, a sales forecasting model is constructed, and the pre-processed multi-dimensional data is input into the sales forecasting model to obtain the sales terminal predicted sales. The specific steps are as follows: The Transformer model is used as the basic model; The input layer accepts preprocessed historical multi-dimensional data for embedding and serialization; The multi-head self-attention layer captures the time series features between different time steps in the sequence; The feed-forward network layer performs nonlinear transformation on each position to extract high-level features; The residual connection layer adds residual connection and layer normalization after each layer; The feature aggregation layer aggregates the output time series features and high-level features into a global representation; The fully connected layer generates prediction values ​​by linearly mapping the aggregated features; Finally, a sales forecasting model was constructed; The preprocessed multi-dimensional data is input into the sales forecasting model to obtain the sales terminal forecast sales volume. The expression is: ; in, It is Sales terminals predict sales volume, It is the first layer after the multi-head self-attention layer. The feature representation of time steps is is the first layer weight matrix of the feedforward neural network, is the bias term of the first layer of the feedforward neural network, is the second layer weight matrix of the feedforward neural network, is the bias term of the second layer of the feedforward neural network, is the attention weight matrix of the time step aggregation, is the attention bias term for time step aggregation, is the weight matrix of the fully connected layer, is the bias term of the fully connected layer, is the total number of time steps, is the index variable of the time step, is the index variable of the sales terminal, is the Sigmoid activation function, is the layer normalization operation, is the ReLU activation function.

[0009] As a preferred solution of the commodity distribution management method based on big data analysis of the present invention, the distribution network graph structure is constructed based on the sales terminal predicted sales volume, and the specific steps are as follows: The predicted sales volume of each sales terminal is mapped into a node in the network graph by means of data structure mapping; Through the set construction method, all mapped nodes are organized into node sets; According to the node and nodes The sales correlation, geographical distance and distribution cost between nodes are used to comprehensively calculate the edge weights between nodes. The expression is: ; in, Representation Node and nodes The edge weights between is the weight coefficient of sales volume correlation, Is a node and nodes The sales correlation between is the weight coefficient of geographical distance, Is a node and nodes The geographical distance between is the weight coefficient of distribution cost, Is a node and nodes The distribution costs between is the first node in the network graph, is the second node in the network graph; According to the edge weights, construct the adjacency matrix; Based on the node set and adjacency matrix, the distribution network graph structure is constructed.

[0010] As a preferred solution of the commodity distribution management method based on big data analysis described in the present invention, wherein: based on the distribution network graph structure, the distribution plan is obtained through a dynamic graph neural network, and the specific steps are as follows: The node embedding in the distribution network graph structure is dynamically optimized through a dynamic graph neural network, and the expression is: ; in, is the time step The optimized node embedding of is the time step The adjacency matrix of is the time step All nodes of is the time step The trainable weight matrix, is the time step The bias term of Based on the optimized node embedding, the distribution optimization target value is calculated, and the expression is: ; in, is the distribution optimization target value, is a hyperparameter that controls the strength of interactions between node embeddings, Is a node At time step The embedding Is a node At time step The embedding is the regularization parameter, Is a node and nodes The distribution costs between is the total number of nodes in the network; Based on the distribution optimization target value, the edge weights are adjusted by gradient descent method; According to the adjusted edge weights, the optimal path between nodes is extracted to obtain the distribution plan.

[0011] As a preferred solution of the commodity distribution management method based on big data analysis described in the present invention, the distribution plan is dynamically adjusted through multi-objective optimization to generate the optimal distribution plan. The specific steps are as follows: The distribution efficiency of the distribution plan is dynamically adjusted through a multi-objective optimization algorithm, and the expression is: ; in, is to maximize distribution efficiency; The distribution cost of the distribution plan is dynamically adjusted through a multi-objective optimization algorithm, and the expression is: ; in, is to minimize distribution costs; Through the weighted sum method, the adjusted maximum distribution efficiency and the adjusted minimum distribution cost are combined to obtain the optimal distribution plan.

[0012] As a preferred solution of the commodity distribution management method based on big data analysis described in the present invention, the specific steps of customizing the regional distribution plan based on the optimal distribution plan through regional adaptive optimization are as follows: Based on the optimal distribution plan, the global network is divided into different regions through clustering algorithms according to the scale, characteristics and business needs of the distribution network; For each region, extract the logistics connections of the sales terminals in the region and generate the corresponding regional sub-network; Optimize and adjust the distribution efficiency and distribution cost of each regional sub-network through multi-objective optimization; After the optimization is completed, the distribution plan within the region is extracted and the regional distribution plan is generated.

[0013] In a second aspect, the present invention provides a commodity distribution management system based on big data analysis, including a data collection and preprocessing module, a sales forecasting module, a distribution network diagram construction module, a distribution plan generation module, a multi-objective adjustment module and a regional adaptive customization module; the data collection and preprocessing module is used to collect multi-dimensional data and preprocess the multi-dimensional data; the sales forecasting module is used to build a sales forecasting model based on a Transformer model, input the preprocessed multi-dimensional data into the sales forecasting model, and obtain the predicted sales volume of the sales terminal; the distribution network diagram construction module is used to build a distribution network diagram structure based on the predicted sales volume of the sales terminal; the distribution plan generation module is used to obtain the distribution plan based on the distribution network diagram structure through a dynamic graph neural network; the multi-objective adjustment module is used to dynamically adjust the distribution plan through multi-objective optimization to generate an optimal distribution plan; the regional adaptive customization module is used to customize the regional distribution plan through regional adaptive optimization based on the optimal distribution plan.

[0014] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the commodity distribution management method based on big data analysis as described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the commodity distribution management method based on big data analysis as described in the first aspect of the present invention.

[0016] The beneficial effects of the present invention are as follows: the present invention solves the problems of insufficient multi-dimensional factor analysis and dynamic adjustment of distribution networks in the prior art through the integration and preprocessing of multi-dimensional data, sales volume forecasting based on the Transformer model, dynamic optimization of distribution network diagrams, and generation of optimal distribution plans through multi-objective optimization. The present invention realizes the full process intelligence of distribution management from demand forecasting to distribution plan formulation, significantly improves distribution efficiency, flexibility and accuracy, and reduces distribution costs, adapts to complex and dynamic market environments, and meets the needs of refined supply chain management. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0018] Figure 1 This is a flow chart of the commodity distribution management method based on big data analysis in Example 1.

[0019] Figure 2 This is a system diagram of the commodity distribution management system based on big data analysis in Example 1. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0021] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0023] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a commodity distribution management method based on big data analysis, comprising the following steps: S1. Multi-dimensional data includes product attribute data, sales terminal data, market environment data, consumer behavior data and distribution cost data.

[0024] S1.1. Preprocessing of multidimensional data includes denoising, normalization, time series processing, feature encoding and feature dimensionality reduction.

[0025] It should be noted that the data is denoised by filtering algorithms, outlier detection or noise removal methods to eliminate random interference and outliers in the data and improve data quality; The data is normalized through methods such as maximum and minimum normalization and Z-score standardization to unify the dimensions and ranges of different features and reduce the impact of scale differences on model calculations; Through methods such as sliding windows, time series segmentation or differential processing, time series data is sorted and analyzed to extract time-dependent features and capture the time series trends and laws of data; Through One-Hot Encoding, Label Encoding or Embedding, categorical or non-numeric features are converted into numerical form, so that non-numeric data can be effectively recognized and processed by the model; The data is reduced in dimension through principal component analysis (PCA), linear discriminant analysis (LDA) or feature selection algorithms to reduce redundant features and computational complexity while retaining the main information of the data to improve model performance.

[0026] S2. Based on the Transformer model, a sales forecasting model is constructed, and the preprocessed multi-dimensional data is input into the sales forecasting model to obtain the sales terminal predicted sales volume.

[0027] S2.1, based on the Transformer model; The input layer accepts preprocessed historical multi-dimensional data for embedding and serialization; The multi-head self-attention layer captures the time series features between different time steps in the sequence; The feed-forward network layer performs nonlinear transformation on each position to extract high-level features; The residual connection layer adds residual connection and layer normalization after each layer; The feature aggregation layer aggregates the output time series features and high-level features into a global representation; The fully connected layer generates prediction values ​​by linearly mapping the aggregated features; Finally, a sales forecasting model was constructed.

[0028] It should be noted that the sales forecasting model is built based on the Transformer model. The Transformer's multi-head self-attention mechanism can capture the complex dependencies between different time steps in the time series and between multi-dimensional features. At the same time, its parallel computing capability significantly improves the processing efficiency of large-scale data. Through residual connection and layer normalization, the model maintains the stability of training during deep feature extraction and avoids the gradient vanishing problem.

[0029] S2.2. Input the preprocessed multi-dimensional data into the sales forecasting model to obtain the sales terminal forecast sales volume, the expression is: ; in, It is Sales terminals predict sales volume, It is the first layer after the multi-head self-attention layer. The feature representation of time steps is is the first layer weight matrix of the feedforward neural network, is the bias term of the first layer of the feedforward neural network, is the second layer weight matrix of the feedforward neural network, is the bias term of the second layer of the feedforward neural network, is the attention weight matrix of the time step aggregation, is the attention bias term for time step aggregation, is the weight matrix of the fully connected layer, is the bias term of the fully connected layer, is the total number of time steps, is the index variable of the time step, is the index variable of the sales terminal, is the Sigmoid activation function, is the layer normalization operation, is the ReLU activation function.

[0030] S3. Construct a distribution network graph structure based on sales terminal forecast sales.

[0031] S3.1. Map the predicted sales volume of each sales terminal into a node in the network graph through data structure mapping.

[0032] S3.2. Organize all mapped nodes into node sets through set construction.

[0033] It should be noted that, through the set construction method, all nodes mapped based on the sales terminal forecast sales are uniformly collected according to the graph theory modeling rules and organized into a node set containing all nodes.

[0034] S3.3, according to the node and nodes The sales correlation, geographical distance and distribution cost between nodes are used to comprehensively calculate the edge weights between nodes. The expression is: ; in, Representation Node and nodes The edge weights between is the weight coefficient of sales volume correlation, Is a node and nodes The sales correlation between is the weight coefficient of geographical distance, Is a node and nodes The geographical distance between is the weight coefficient of distribution cost, Is a node and nodes The distribution costs between is the first node in the network graph, is the second node in the network graph.

[0035] It should be noted that the geographical distance is calculated using the latitude and longitude coordinates of the sales terminal using the spherical distance formula or the Euclidean distance; Distribution costs are calculated based on unit transportation costs and geographical distance, combined with the company's internal logistics data or quotations from logistics service providers.

[0036] S3.4. Construct an adjacency matrix based on edge weights.

[0037] It should be noted that the calculated edge weight values ​​are filled into the corresponding positions in a two-dimensional table. The rows and columns of this table correspond to the nodes in the graph; For each edge between a pair of nodes, if there is a connection, fill in the weight value of the edge at the intersection of the corresponding row and column in the table; if there is no direct connection, it is usually filled in with 0 (or infinity, depending on the application scenario); For undirected graphs, since edges have no direction, the table is symmetric, i.e., the weight from node A to node B is equal to the weight from node B to node A, whereas for directed graphs, the table may not be symmetric, since edge weights in different directions can be different; The final table fully describes the connection strength and relationship between nodes in the graph, thereby constructing an adjacency matrix.

[0038] S3.5. Based on the node set and adjacency matrix, construct the distribution network graph structure.

[0039] It should be noted that the distribution network graph structure is constructed by using graph theory modeling, using node sets and adjacency matrices, taking nodes as graph vertices, and non-zero edge weights in the adjacency matrix as connection relationships between vertices.

[0040] S4. Based on the distribution network graph structure, the distribution plan is obtained through dynamic graph neural network.

[0041] S4.1. Dynamically optimize the node embedding in the distribution network graph structure through a dynamic graph neural network. The expression is: ; in, is the time step The optimized node embedding of is the time step The adjacency matrix of is the time step All nodes of is the time step The trainable weight matrix, is the time step The bias term.

[0042] It should be noted that the specific process of node embedding is: When constructing a distribution network graph structure based on sales terminal sales forecasts, each sales terminal is converted into a node in the graph, and the edge weights between nodes are calculated based on factors such as sales correlation, geographical distance, and distribution cost to form node embedding; It should also be noted that the specific process of dynamic optimization is: Dynamic graph neural networks are used to dynamically optimize these node embeddings. That is, as time or other factors change, the node embeddings are iteratively updated by adjusting parameters such as weight matrices and bias terms to ensure that they reflect the latest network status, thereby optimizing distribution paths and plans.

[0043] S4.2. Based on the optimized node embedding, calculate the distribution optimization target value, the expression is: ; in, is the distribution optimization target value, is a hyperparameter that controls the strength of interactions between node embeddings, Is a node At time step The embedding Is a node At time step The embedding is the regularization parameter, Is a node and nodes The distribution costs between is the total number of nodes in the network.

[0044] It should be noted that the characters from becomes and This is because a subscript is introduced in the expression to indicate that a specific node is at a certain time step. The embedding state of the node at a specific time step is emphasized.

[0045] S4.3. Based on the distribution optimization target value, adjust the edge weights through the gradient descent method.

[0046] It should be noted that by calculating the gradient of the objective function with respect to the edge weight, the adjustment direction and amplitude of the current edge weight are determined, and the edge weight is iteratively optimized according to the update rule of the gradient descent method, so that the value of the objective function gradually approaches the optimal solution. Through multiple iterations, the edge weight configuration that meets the distribution optimization goal is obtained.

[0047] S4.4. According to the adjusted edge weights, the optimal path between nodes is extracted to obtain the distribution plan.

[0048] It should be noted that the weighted graph structure of the distribution network is constructed using the adjusted edge weights, where the edge weights reflect the distribution cost or association strength between nodes. The distribution starting point (such as a warehouse node) is used as the source node, and the optimal path to each target node (such as a sales terminal) is calculated through the shortest path algorithm (such as the Dijkstra algorithm or the A* algorithm); then the distribution relationship between the nodes is extracted based on these paths, and all paths are combined to generate an optimal distribution plan covering all target nodes.

[0049] S5. Dynamically adjust the distribution plan through multi-objective optimization to generate the optimal distribution plan.

[0050] S5.1. Dynamically adjust the distribution efficiency of the distribution plan through a multi-objective optimization algorithm, expressed as: ; in, is to maximize distribution efficiency; S5.2. Dynamically adjust the distribution cost of the distribution plan through a multi-objective optimization algorithm, expressed as: ; in, is to minimize distribution costs.

[0051] S5.3. By using the weighted sum method, the adjusted maximum distribution efficiency and the adjusted minimum distribution cost are combined to obtain the optimal distribution plan.

[0052] It should be noted that the optimal distribution plan is obtained by weighted sum method, and the expression is: ; in, is the best trade-off value, is the weight of distribution efficiency, is the weight of distribution cost; It should also be noted that the weighted sum method is a multi-objective optimization method that combines multiple optimization objectives into a single objective function by assigning weights to them.

[0053] S6. Based on the optimal distribution plan, customize the regional distribution plan through regional adaptive optimization.

[0054] S6.1. Based on the optimal distribution plan, the global network is divided into different regions through clustering algorithm according to the scale, characteristics and business needs of the distribution network.

[0055] It should be noted that the global network is a distribution network graph structure built based on the predicted sales of sales terminals. It maps the predicted sales of sales terminals, geographical distances and distribution costs into nodes and edges in the network graph, defines edge weights, and then uses the adjacency matrix and graph theory modeling methods to generate a network graph. It reflects the relationship between all nodes and paths in the entire distribution system and is the core of achieving distribution optimization from global to regional.

[0056] S6.2. For each region, extract the logistics connections of the sales terminals in the region and generate the corresponding regional sub-network.

[0057] It should be noted that according to the areas divided by the clustering algorithm, the sales terminal nodes belonging to the same area are screened out, and the connection relationships between these nodes and the corresponding edge weights are extracted from the adjacency matrix or edge set of the global network. Based on the extracted nodes and edges, a weighted subnetwork covering only the sales terminals in the area and their logistics connections is constructed.

[0058] S6.3. Optimize and adjust the distribution efficiency and distribution cost of each regional sub-network through multi-objective optimization.

[0059] S6.4. After the optimization is completed, the distribution plan within the region is extracted and the regional distribution plan is generated.

[0060] Specifically, based on the optimization results of the regional sub-network, the optimal distribution path and logistics allocation plan for each sales terminal in the region are determined, the logistics connection between nodes and the corresponding distribution volume or cost information are extracted from the optimized sub-network, and these paths and distribution configurations are integrated to generate a specific distribution plan suitable for the region.

[0061] This embodiment also provides a commodity distribution management system based on big data analysis, including: a data collection and preprocessing module, a sales volume forecasting module, a distribution network diagram construction module, a distribution plan generation module, a multi-objective adjustment module and a regional adaptive customization module; Data collection and preprocessing module, used to collect multi-dimensional data and preprocess the multi-dimensional data; The sales forecasting module is used to build a sales forecasting model based on the Transformer model, input the pre-processed multi-dimensional data into the sales forecasting model, and obtain the sales terminal predicted sales; A distribution network diagram construction module is used to construct a distribution network diagram structure based on sales terminal forecast sales; The distribution plan generation module is used to obtain the distribution plan based on the distribution network graph structure through a dynamic graph neural network; Multi-objective adjustment module, used to dynamically adjust the distribution plan through multi-objective optimization to generate the optimal distribution plan; The regional adaptive customization module is used to customize the regional distribution plan through regional adaptive optimization based on the optimal distribution plan.

[0062] This embodiment also provides a computer device, which is suitable for the commodity distribution management method based on big data analysis, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the commodity distribution management method based on big data analysis proposed in the above embodiment.

[0063] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0064] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for implementing the commodity distribution management method based on big data analysis proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0065] In summary, the present invention solves the problems of insufficient multi-dimensional factor analysis and dynamic adjustment of distribution networks in the prior art through: integration and preprocessing of multi-dimensional data, sales forecasting based on the Transformer model, dynamic optimization of distribution network diagrams, and generation of optimal distribution plans through multi-objective optimization. It realizes the intelligence of the entire process of distribution management from demand forecasting to distribution plan formulation, significantly improves distribution efficiency, flexibility and accuracy, and reduces distribution costs, adapts to complex and dynamic market environments, and meets the needs of refined supply chain management.

[0066] Example 2, referring to Table 1, is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of a commodity distribution management method based on big data analysis is provided.

[0067] In order to verify the effectiveness and innovation of the intelligent distribution solution based on multi-dimensional data, a large electronic product retailer was selected as the research object. The retailer has more than 100 sales terminals across the country and has accumulated a large amount of commodity attribute data, sales terminal data, market environment data, consumer behavior data and distribution cost data in the past few years. This experiment aims to improve the inventory management and logistics distribution efficiency of retailers by introducing a new sales forecasting model and distribution network graph structure construction method.

[0068] First, the experiment preprocessed the collected multi-dimensional data such as product attributes, sales terminals, market environment, consumer behavior and distribution costs, including denoising, normalization, time series analysis, feature encoding and dimensionality reduction, to improve data quality and applicability.

[0069] Secondly, a sales forecasting model was built based on the Transformer model. The model was trained using preprocessed multi-dimensional data to accurately obtain the predicted sales of each sales terminal, and a distribution network graph structure reflecting the logistics relationship was constructed based on this.

[0070] Next, the distribution network was optimized through a dynamic graph neural network, the edge weights were adjusted to find the optimal path, and a multi-objective optimization algorithm was used to balance distribution efficiency and cost to generate an efficient distribution plan.

[0071] Finally, the experimental results show that the new method is significantly superior to the traditional method in terms of sales forecast accuracy, distribution cost control, logistics efficiency and customer satisfaction, proving the effectiveness and innovation of the intelligent distribution solution.

[0072] The traditional method refers to sales forecasting and distribution network planning based on historical data and experience, relying on simple statistical models and manual decision-making.

[0073] The details are shown in Table 1 below: Table 1 Comparison of sales forecast and distribution optimization effects

[0074] Through the data analysis of the above table, it can be clearly seen that the method of the present invention has shown significant advantages over the traditional method in multiple key indicators. The present invention optimizes the sales forecast and distribution network construction process by introducing advanced multi-dimensional data analysis and intelligent algorithms. For example, in terms of sales forecast error rate, the error rate of the traditional method is 14.2%, while the method of the present invention reduces this value to 4.8%, greatly improving the prediction accuracy; in terms of distribution cost saving rate, the present invention achieves 25.3% cost savings, which is much higher than the 9.8% of the traditional method, effectively reducing the operating costs of the enterprise. The present invention also significantly improves logistics efficiency to 20.4%, compared with only 6.1% for the traditional method, which shows that the new method can respond to changes in market demand more quickly and improve customer satisfaction scores to 4.5 points (out of 5 points), which is a significant improvement over the 3.9 points of the traditional method. At the same time, the data processing time is shortened from an average of 75 seconds to 35 seconds, which speeds up decision-making, and the inventory turnover rate is increased by 15.2%, and the out-of-stock rate is reduced by 11.8%, further proving the efficiency and reliability of the invention in practical applications. These improvements not only enhance the company's competitiveness, but also provide customers with a smoother service experience.

[0075] The present invention significantly improves the accuracy of sales forecasts, significantly reduces distribution costs, optimizes logistics efficiency, and shortens data processing time by introducing advanced multi-dimensional data analysis and intelligent algorithms, thereby achieving more efficient inventory management and faster market response, ultimately enhancing the competitiveness of the enterprise and providing a better and smoother customer experience.

[0076] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A commodity distribution management method based on big data analysis, characterized in that: include, Collect multi-dimensional data and pre-process the multi-dimensional data; Based on the Transformer model, a sales forecasting model is constructed, and the pre-processed multi-dimensional data is input into the sales forecasting model to obtain the sales terminal forecasted sales; Construct a distribution network graph structure based on sales terminal forecast sales; Based on the distribution network graph structure, the distribution plan is obtained through dynamic graph neural network; Dynamically adjust the distribution plan through multi-objective optimization to generate the optimal distribution plan; Based on the optimal distribution plan, the regional distribution plan is customized through regional adaptive optimization.

2. The commodity distribution management method based on big data analysis according to claim 1, characterized in that: The multi-dimensional data includes commodity attribute data, sales terminal data, market environment data, consumer behavior data and distribution cost data; The multi-dimensional data preprocessing includes denoising, normalization, time series processing, feature encoding and feature dimensionality reduction.

3. The commodity distribution management method based on big data analysis according to claim 2, characterized in that: Based on the Transformer model, a sales forecasting model is constructed, and the pre-processed multi-dimensional data is input into the sales forecasting model to obtain the sales terminal predicted sales. The specific steps are as follows: The Transformer model is used as the basic model; The input layer accepts preprocessed historical multi-dimensional data for embedding and serialization; The multi-head self-attention layer captures the time series features between different time steps in the sequence; The feed-forward network layer performs nonlinear transformation on each position to extract high-level features; The residual connection layer adds residual connection and layer normalization after each layer; The feature aggregation layer aggregates the output time series features and high-level features into a global representation; The fully connected layer generates prediction values ​​by linearly mapping the aggregated features; Finally, a sales forecasting model was constructed; The preprocessed multi-dimensional data is input into the sales forecasting model to obtain the sales terminal forecast sales volume. The expression is: ； in, It is Sales terminals predict sales volume, It is the first layer after the multi-head self-attention layer. The feature representation of time steps is is the first layer weight matrix of the feedforward neural network, is the bias term of the first layer of the feedforward neural network, is the second layer weight matrix of the feedforward neural network, is the bias term of the second layer of the feedforward neural network, is the attention weight matrix of the time step aggregation, is the attention bias term for time step aggregation, is the weight matrix of the fully connected layer, is the bias term of the fully connected layer, is the total number of time steps, is the index variable of the time step, is the index variable of the sales terminal, is the Sigmoid activation function, is the layer normalization operation, is the ReLU activation function.

4. The commodity distribution management method based on big data analysis according to claim 3, characterized in that: The specific steps of constructing a distribution network graph structure based on sales terminal sales forecasting are as follows: The predicted sales volume of each sales terminal is mapped into a node in the network graph by means of data structure mapping; Through the set construction method, all mapped nodes are organized into node sets; According to the node and nodes The sales correlation, geographical distance and distribution cost between nodes are used to comprehensively calculate the edge weights between nodes. The expression is: ； in, Representation Node and nodes The edge weights between is the weight coefficient of sales volume correlation, Is a node and nodes The sales correlation between is the weight coefficient of geographical distance, Is a node and nodes The geographical distance between is the weight coefficient of distribution cost, Is a node and nodes The distribution costs between is the first node in the network graph, is the second node in the network graph; According to the edge weights, construct the adjacency matrix; Based on the node set and adjacency matrix, the distribution network graph structure is constructed.

5. The commodity distribution management method based on big data analysis according to claim 4, characterized in that: Based on the distribution network graph structure, the distribution plan is obtained through a dynamic graph neural network. The specific steps are as follows: The node embedding in the distribution network graph structure is dynamically optimized through a dynamic graph neural network, and the expression is: ； in, is the time step The optimized node embedding of is the time step The adjacency matrix of is the time step All nodes of is the time step The trainable weight matrix, is the time step The bias term of Based on the optimized node embedding, the distribution optimization target value is calculated, and the expression is: ； in, is the distribution optimization target value, is a hyperparameter that controls the strength of interactions between node embeddings, Is a node At time step The embedding Is a node At time step The embedding is the regularization parameter, Is a node and nodes The distribution costs between is the total number of nodes in the network; Based on the distribution optimization target value, the edge weights are adjusted by gradient descent method; According to the adjusted edge weights, the optimal path between nodes is extracted to obtain the distribution plan.

6. The commodity distribution management method based on big data analysis according to claim 5, characterized in that: The distribution plan is dynamically adjusted through multi-objective optimization to generate the optimal distribution plan. The specific steps are as follows: The distribution efficiency of the distribution plan is dynamically adjusted through a multi-objective optimization algorithm, and the expression is: ； in, is to maximize distribution efficiency; The distribution cost of the distribution plan is dynamically adjusted through a multi-objective optimization algorithm, and the expression is: ； in, is to minimize distribution costs; Through the weighted sum method, the adjusted maximum distribution efficiency and the adjusted minimum distribution cost are combined to obtain the optimal distribution plan.

7. The commodity distribution management method based on big data analysis according to claim 6, characterized in that: Based on the optimal distribution plan, the regional distribution plan is customized through regional adaptive optimization. The specific steps are as follows: Based on the optimal distribution plan, the global network is divided into different regions through clustering algorithms according to the size, characteristics and business needs of the distribution network; For each region, extract the logistics connections of the sales terminals in the region and generate the corresponding regional sub-network; Optimize and adjust the distribution efficiency and distribution cost of each regional sub-network through multi-objective optimization; After the optimization is completed, the distribution plan within the region is extracted and the regional distribution plan is generated.

8. A commodity distribution management system based on big data analysis, based on the commodity distribution management method based on big data analysis according to any one of claims 1 to 7, characterized in that: It includes data collection and preprocessing module, sales forecasting module, distribution network diagram building module, distribution plan generation module, multi-objective adjustment module and regional adaptive customization module; Data collection and preprocessing module, used to collect multi-dimensional data and preprocess the multi-dimensional data; The sales forecasting module is used to build a sales forecasting model based on the Transformer model, input the pre-processed multi-dimensional data into the sales forecasting model, and obtain the sales terminal predicted sales; A distribution network diagram construction module is used to construct a distribution network diagram structure based on sales terminal forecast sales; The distribution plan generation module is used to obtain the distribution plan based on the distribution network graph structure through a dynamic graph neural network; Multi-objective adjustment module, used to dynamically adjust the distribution plan through multi-objective optimization to generate the optimal distribution plan; The regional adaptive customization module is used to customize the regional distribution plan through regional adaptive optimization based on the optimal distribution plan.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the commodity distribution management method based on big data analysis described in any one of claims 1 to 7 are implemented.

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 steps of the commodity distribution management method based on big data analysis described in any one of claims 1 to 7 are implemented.

Images