Logistics supply chain dynamic management system and method based on big data

Abstract

The invention relates to the technical field of logistics supply chain management, and discloses a logistics supply chain dynamic management system and method based on big data, a big data acquisition module is used for acquiring data of each link of a logistics supply chain; the demand prediction module is used for predicting the preprocessed customer demand data by adopting an improved long-short term memory (LSTM) network, namely an LSTM demand prediction model; and the inventory optimization module uses an improved economic order quantity model, namely an EOQ model, to calculate an optimal order quantity and an optimal order period according to the demand prediction value, the inventory data and the supplier data, and the logistics supply chain dynamic management system and method based on the big data realize intelligent management of each link of the logistics supply chain. The dynamic monitoring module can find abnormal conditions in real time and feed back the abnormal conditions in time, the decision support module can generate scientific and reasonable decision suggestions, and the response speed and management efficiency of the logistics supply chain are improved.

Description

Technical Field

[0001] This invention relates to the field of logistics supply chain management technology, specifically to a dynamic management system and method for logistics supply chain based on big data. Background Technology

[0002] With the rapid development of e-commerce and the formation of a globalized market, the scale of logistics supply chains is constantly expanding, and their structure is becoming increasingly complex, placing higher demands on logistics supply chain management. Traditional logistics supply chain management systems mainly rely on human experience for decision-making, which suffers from problems such as low accuracy in demand forecasting, insufficient flexibility in route planning, and poor dynamics in inventory optimization. These systems struggle to adapt to complex and ever-changing market environments and customer needs, resulting in low operational efficiency and high costs in the logistics supply chain.

[0003] In terms of demand forecasting, traditional methods such as moving averages and exponential smoothing only consider the linear trend of historical data and cannot effectively handle nonlinear and non-stationary customer demand data. As a result, the forecasting accuracy is low and it is easy to cause inventory backlog or stockouts.

[0004] In terms of route planning, traditional route planning methods such as Dijkstra's algorithm and Floyd's algorithm usually only consider a single objective, such as the lowest transportation cost or the shortest transportation time, ignoring other important factors such as transportation risks, and are difficult to cope with changes in real-time road conditions, resulting in poor flexibility and practicality in route planning.

[0005] In terms of inventory optimization, the traditional Economic Order Quantity (EOQ) model assumes that demand is constant and supply is timely, without considering the impact of demand fluctuations and supply delays. This leads to a large deviation between the calculated optimal order quantity and order cycle and the actual situation, making it impossible to achieve dynamic inventory optimization.

[0006] To address the above problems, this invention proposes a dynamic management system and method for logistics supply chain based on big data. Summary of the Invention

[0007] This invention provides a dynamic management system and method for logistics supply chain based on big data, which helps to solve the problems mentioned in the background art.

[0008] This invention provides the following technical solution: a big data-based dynamic logistics supply chain management system, comprising a big data acquisition module, a data preprocessing module, a demand forecasting module, a route planning module, an inventory optimization module, a dynamic monitoring module, and a decision support module.

[0009] The big data acquisition module is used to collect data from all aspects of the logistics supply chain, including customer demand data, transportation data, inventory data, supplier data, and market environment data.

[0010] The data preprocessing module is used to clean, integrate, transform, and reduce the collected raw data to obtain a standardized dataset.

[0011] The demand forecasting module uses an improved Long Short-Term Memory (LSTM) network demand forecasting model to predict preprocessed customer demand data and output predicted customer demand values ​​for a future period of time.

[0012] The route planning module, based on an improved ant colony algorithm, combines transportation data and real-time traffic data to construct a multi-objective route planning model and output the optimal transportation route.

[0013] The inventory optimization module uses an improved economic order quantity (EOQ) model to calculate the optimal order quantity and order cycle based on demand forecasts, inventory data, and supplier data.

[0014] The dynamic monitoring module collects operational data from each link of the logistics supply chain in real time, monitors the transportation process, inventory levels, and order execution, and feeds back abnormal information to the decision support module. The decision support module generates decision suggestions based on demand forecast results, optimal transportation routes, optimal inventory plans, and abnormal information, providing support for logistics supply chain management.

[0015] Furthermore, the output end of the big data acquisition module is electrically connected to the input end of the data preprocessing module, and the output end of the data preprocessing module is electrically connected to the input end of the demand forecasting module. The improved LSTM demand forecasting model in the demand forecasting module introduces an attention mechanism and a time decay factor on the basis of the LSTM model. The attention mechanism is used to assign different weights to the input data at different time steps, and the time decay factor is used to adjust the degree of influence of historical data.

[0016] The formula for calculating the output layer of the improved LSTM demand forecasting model is as follows: Where y t W represents the demand forecast at time t. o b is the output layer weight matrix; o This is the output bias vector; The cell state after adjustment by attention mechanisms and time decay factors;

[0017] The weight calculation formula for the attention mechanism is as follows: Where α t,i Let e ​​be the attention weight for the input data at time t to the i-th time step; t,i A score for attention; v a Wa U a b is the parameter matrix of the attention mechanism; a h is the bias vector for the attention mechanism. i h represents the hidden state at the i-th time step. t Let be the hidden state at time t; T is the total number of time steps.

[0018] The formula for calculating the time decay factor is: λ t,i =exp(-β·(ti)), where λ t,i β is the time decay factor for the input data at time t at the i-th time step; β is the decay coefficient, and its value ranges from 0.01 to 0.1.

[0019] Furthermore, the multi-objective route planning model takes the minimum transportation cost, the shortest transportation time, and the minimum transportation risk as its objective functions. The expression of the objective function is: minF=(ω1·C,ω2·T′,ω3·R), where F is the multi-objective optimization function; ω1, ω2, and ω3 are the weight coefficients of transportation cost, transportation time, and transportation risk, respectively, and ω1+ω2+ω3=1; C is the transportation cost; T′ is the transportation time; and R is the transportation risk.

[0020] The calculation strategy for the transportation cost C is as follows: Where n is the total number of nodes; d i,j Let be the distance from node i to node j; c is the transportation cost per unit distance; x i,j It is a 0-1 variable, and x i,j =1 indicates that there is a transportation path from node i to node j, x i,j =0 indicates no transport path; f i y represents the fixed cost of node i; i It is a 0-1 variable, and y i =1 indicates that node i is a docking point, y i =0 indicates no selection;

[0021] The calculation strategy for the transportation time T′ is as follows: Where v i,j Let t' be the transport speed from node i to node j; i Let be the dwell time at node i;

[0022] The calculation strategy for the transportation risk R is as follows: Where r i,j Let be the transportation risk coefficient from node i to node j; it is determined based on road conditions, weather, and traffic congestion.

[0023] Furthermore, the improved ant colony algorithm in the path planning module introduces an adaptive pheromone update strategy and a local search operator on the basis of the ant colony algorithm. The adaptive pheromone update strategy specifically involves dynamically adjusting the pheromone evaporation coefficient and increment according to the quality of the path. The local search operator is used to optimize the initial solution obtained by the ant colony algorithm.

[0024] The improved pheromone update formula for the ant colony algorithm is as follows: Where τ i,j (t+1) represents the pheromone concentration from node i to node j at time t+1; ρ(t) represents the pheromone volatility coefficient at time t; τ i,j (t) represents the pheromone concentration from node i to node j at time t; m represents the number of ants in the colony. Let be the pheromone increment of the k-th ant on the path from node i to node j at time t;

[0025] The adaptive adjustment formula for the pheromone volatile coefficient ρ(t) is as follows: Where ρ min ρ is the minimum pheromone evaporation coefficient. max γ is the maximum pheromone evaporation coefficient; γ is the adjustment coefficient; iter is the current iteration number; iter max This represents the maximum number of iterations.

[0026] The pheromone increment The calculation formula is: Where Q represents the total amount of pheromones; F k The value of the multi-objective optimization function for the path taken by the k-th ant.

[0027] Furthermore, the improved EOQ model in the inventory optimization module introduces a safety stock factor and a supply delay coefficient based on the EOQ model. The optimal order quantity calculation strategy of the improved EOQ model is as follows: Q * The optimal order quantity is σ; D is the annual demand; S is the cost per order; H is the annual holding cost per unit of inventory; σ is the minimum order quantity. D The standard deviation of demand; μ D The mean demand is represented by z; the safety stock factor is determined based on the service level; and L is the supply delay time.

[0028] The optimal order cycle calculation strategy of the improved EOQ model is as follows: Where T * This is the optimal ordering cycle.

[0029] Furthermore, a dynamic management method for logistics supply chain based on big data includes the following steps:

[0030] S1: Data Acquisition. Data from all aspects of the logistics supply chain is collected through the big data acquisition module, including customer demand data, transportation data, inventory data, supplier data, and market environment data.

[0031] S2: Data preprocessing. The data preprocessing module cleans the collected raw data, removing outliers and missing values; integrates the cleaned data, consolidating data from different sources into a unified data warehouse; transforms the integrated data into a format suitable for model analysis; and reduces the dimensionality and quantity of the transformed data to obtain a standardized dataset.

[0032] S3: Demand Forecasting. The demand forecasting module uses an improved LSTM demand forecasting model to forecast customer demand data in a standardized dataset.

[0033] S31. Normalize the customer demand data and map the data to the [0,1] range;

[0034] S32. Divide the normalized customer demand data into a training set and a test set according to the time series;

[0035] S33. Train the improved LSTM demand prediction model using the training set, and adjust the model parameters through the backpropagation algorithm to minimize the prediction error of the model;

[0036] S34. Use the test set to validate the trained model. If the model's prediction accuracy meets the preset requirements, output the predicted value of customer demand for a future period of time; otherwise, readjust the model parameters and train it.

[0037] S4: Route planning. The route planning module constructs a multi-objective route planning model based on demand forecasts, transportation data, and real-time traffic data, and uses an improved ant colony algorithm to solve the multi-objective route planning model.

[0038] S41. Initialize the parameters of the ant colony algorithm, including the number of ants, the initial pheromone concentration, and the maximum number of iterations;

[0039] S42. Each ant selects the next node based on the pheromone concentration between nodes and a heuristic function, thus constructing a transportation path;

[0040] S43. Calculate the multi-objective optimization function value of the path taken by each ant, and update the pheromone concentration between nodes according to the adaptive pheromone update strategy;

[0041] S44. Introduce local search operators to optimize the initial solution and improve the quality of the solution;

[0042] S45. When the number of iterations reaches the maximum number of iterations or the quality of the solution meets the preset requirements, output the optimal transportation path;

[0043] S5: Inventory Optimization. The inventory optimization module uses an improved EOQ model to calculate the optimal order quantity and lead time based on demand forecasts, inventory data, and supplier data.

[0044] S51. Obtain parameters such as annual demand, cost per order, annual holding cost per unit of inventory, standard deviation of demand, mean demand, service level, and supply delay time.

[0045] S52 substitutes the parameters obtained in S51 into the optimal order quantity calculation strategy and optimal order cycle calculation strategy of the improved EOQ model to obtain the optimal order quantity and order cycle.

[0046] S6: Dynamic monitoring. The dynamic monitoring module collects operational data from all aspects of the logistics supply chain in real time, including the location, speed, inventory level, and order execution progress of transport vehicles. It analyzes the collected operational data in real time to determine whether the transportation process, inventory level, and order execution are normal. If any abnormalities are found, such as transport vehicles deviating from the optimal route, inventory levels falling below safety stock, or order execution delays, the abnormal information is fed back to the decision support module.

[0047] S7: Decision Support. The decision support module generates decision suggestions based on demand forecast results, optimal transportation routes, optimal inventory plans, and abnormal information, combined with the overall operational goals of the logistics supply chain. These suggestions include adjusting transportation plans, optimizing inventory structure, and negotiating with suppliers to adjust delivery times. The decision suggestions are then presented to logistics supply chain managers in a visual manner to support their decision-making.

[0048] Furthermore, in S3, the normalization process employs the Min-Max normalization method, specifically as follows: Where x′ represents the normalized data, x represents the original data, and x... min x is the minimum value in the original data. max This represents the maximum value in the original data.

[0049] Furthermore, in S4, the heuristic function is specifically: Where η i,j C represents the heuristic function value from node i to node j. i,j Let T′ be the unit transportation cost from node i to node j. i,j R is the unit transport time from node i to node j. i,j The unit transportation risk from node i to node j;

[0050] The local search operator employs a 2-opt swapping strategy, which swaps two nodes in the path and recalculates the multi-objective optimization function value of the path. If the swapped path is better, the new path is retained; otherwise, the original path is retained.

[0051] Furthermore, in S6, the judgment of abnormal situations adopts the threshold judgment method. Threshold ranges are set in advance for each operating indicator. If the operating indicator exceeds the threshold range, it is judged as an abnormal situation.

[0052] The operational metrics include vehicle deviation, inventory levels, and order fulfillment delay time.

[0053] The formula for calculating the deviation of the transport vehicle is: Where D′ is the deviation of the transport vehicle; m′ is the number of monitoring time intervals; d t D'0 represents the distance the transport vehicle deviates from the optimal path during the t-th time interval; D'0 represents the total distance of the optimal path.

[0054] Furthermore, in S7, the visualization display adopts charts including line charts, bar charts, pie charts, and maps. The line chart is used to display demand forecast trends and inventory level change trends, the bar chart is used to display the cost and time comparison of different transportation routes, the pie chart is used to display the inventory structure distribution, and the map is used to display the real-time location and optimal route of the transportation vehicles.

[0055] The present invention has the following beneficial effects:

[0056] This big data-based dynamic management system and methodology for logistics supply chains integrates functions such as big data collection, preprocessing, demand forecasting, route planning, inventory optimization, dynamic monitoring, and decision support, enabling intelligent management of all aspects of the logistics supply chain. The dynamic monitoring module can detect anomalies in real time and provide timely feedback, while the decision support module can generate scientific and reasonable decision suggestions, improving the response speed and management efficiency of the logistics supply chain.

[0057] The improved LSTM demand forecasting model introduces an attention mechanism and a time decay factor, which can effectively capture the nonlinear and non-stationary trends of customer demand. It assigns different weights and degrees of influence to historical data at different time steps, which significantly improves the accuracy of demand forecasting.

[0058] The multi-objective path planning model considers transportation costs, transportation time, and transportation risks simultaneously. The improved ant colony algorithm introduces an adaptive pheromone update strategy and a local search operator, which improves the search efficiency and solution quality of the algorithm. It can dynamically adjust the transportation route according to real-time traffic conditions and business needs, thereby reducing logistics transportation costs.

[0059] The improved EOQ model takes into account the impact of demand fluctuations and supply delays. By introducing a safety stock factor and a supply delay coefficient, the calculated optimal order quantity and order cycle are more in line with reality, enabling dynamic optimization of inventory, reducing inventory backlog rate, lowering capital occupation costs, and avoiding stockouts. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of the system flow of the present invention;

[0061] Figure 2 This is a schematic diagram of the method flow of the present invention.

[0062] In the diagram: 1. Big Data Acquisition Module; 2. Data Preprocessing Module; 3. Demand Forecasting Module; 4. Route Planning Module; 5. Inventory Optimization Module; 6. Dynamic Monitoring Module; 7. Decision Support Module. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] Example 1

[0065] Reference Figure 1 A big data-based dynamic logistics supply chain management system includes a big data acquisition module 1, a data preprocessing module 2, a demand forecasting module 3, a route planning module 4, an inventory optimization module 5, a dynamic monitoring module 6, and a decision support module 7.

[0066] The big data acquisition module 1 is used to collect data from all aspects of the logistics supply chain, including customer demand data, transportation data, inventory data, supplier data, and market environment data. Customer demand data includes customer order quantity, order time, product type, etc.; transportation data includes the location, speed, route, and cost of transport vehicles, etc.; inventory data includes inventory quantity, location, and turnover rate, etc.; supplier data includes supplier supply capacity, delivery time, and cost, etc.; and market environment data includes market prices, market demand trends, and competitor information, etc.

[0067] The data preprocessing module 2 is used to clean, integrate, transform and reduce the collected raw data to obtain a standardized dataset. The cleaning, integration, transformation and reduction of the raw data are all performed using existing technologies, so they will not be described in detail here.

[0068] The demand forecasting module 3 uses an improved Long Short-Term Memory (LSTM) network demand forecasting model to predict the preprocessed customer demand data and output the predicted customer demand value for a future period of time.

[0069] The output of the big data acquisition module 1 is electrically connected to the input of the data preprocessing module 2, and the output of the data preprocessing module 2 is electrically connected to the input of the demand forecasting module 3. The improved LSTM demand forecasting model in the demand forecasting module 3 introduces an attention mechanism and a time decay factor on the basis of the LSTM model. The attention mechanism is used to assign different weights to the input data at different time steps, and the time decay factor is used to adjust the degree of influence of historical data.

[0070] The formula for calculating the output layer of the improved LSTM demand forecasting model is as follows: Where y t W represents the demand forecast at time t. o b is the output layer weight matrix; o This is the output bias vector; The cell state after adjustment by attention mechanisms and time decay factors;

[0071] The weight calculation formula for the attention mechanism is as follows: Where α t,i Let e ​​be the attention weight for the input data at time t to the i-th time step; t,i A score for attention; v a W a U a b is the parameter matrix of the attention mechanism; a h is the bias vector for the attention mechanism. i h represents the hidden state at the i-th time step. t Let t be the hidden state at time t; T is the total number of time steps. By calculating the attention weights of the input data at different time steps, the model can focus on historical data that has a greater impact on current demand during the prediction process, thereby improving the targeting of the prediction. For example, in the lead-up to a major e-commerce promotion, recent customer browsing and shopping data have a much greater impact on demand prediction than regular data from several months ago, and the attention mechanism can automatically assign higher weights to recent data.

[0072] The formula for calculating the time decay factor is: λ t,i =exp(-β·(ti)), where λ t,iβ is the time decay factor for the input data at time t at the i-th time step; β is the decay coefficient, ranging from 0.01 to 0.1. The value of β can be adapted according to the demand characteristics of different industries. For example, the demand of the fast-moving consumer goods industry changes rapidly, so β ​​can be 0.08-0.1, while the demand of the durable goods industry changes relatively slowly, so β ​​can be 0.01-0.03. In the training process of the improved LSTM demand forecasting model, the customer demand data after Min-Max normalization is first divided into training and test sets in a 7:3 ratio. Normalization can eliminate the interference of data magnitude differences on model training. During the training phase, the Adam optimizer is used to minimize the root mean square error (RMSE) of the model. The weight matrix and bias vector in the model are continuously adjusted through the backpropagation algorithm until the RMSE of the model converges on the training set and the prediction accuracy on the test set reaches the preset threshold.

[0073] The path planning module 4, based on the improved ant colony algorithm, combines transportation data and real-time traffic data to construct a multi-objective path planning model and output the optimal transportation path.

[0074] In practical applications, the importance of each objective varies in different scenarios. For example, the transportation of fresh products should prioritize ensuring the shortest transportation time (ω2 can be 0.5-0.6), while the transportation of bulk goods focuses more on minimizing transportation costs (ω1 can be 0.5-0.6), and the transportation of dangerous goods requires a focus on controlling transportation risks (ω3 can be 0.5-0.6). The weighting coefficients ω1, ω2, and ω3 can be flexibly adjusted according to actual business needs, and always satisfy ω1+ω2+ω3=1.

[0075] The multi-objective route planning model takes the minimum transportation cost, the shortest transportation time, and the minimum transportation risk as its objective functions. The expression of the objective function is: minF=(ω1·C,ω2·T′,ω3·R), where F is the multi-objective optimization function; ω1, ω2, and ω3 are the weight coefficients of transportation cost, transportation time, and transportation risk, respectively, and ω1+ω2+ω3=1; C is the transportation cost; T′ is the transportation time; and R is the transportation risk.

[0076] The calculation strategy for the transportation cost C is as follows: Where n is the total number of nodes; d i,j Let be the distance from node i to node j; c is the transportation cost per unit distance; x i,j It is a 0-1 variable, and x i,j =1 indicates that there is a transportation path from node i to node j, x i,j =0 indicates no transport path; f i y represents the fixed cost of node i; i It is a 0-1 variable, and y i =1 indicates that node i is a docking point, yi =0 indicates no selection;

[0077] The calculation strategy for the transportation time T′ is as follows: Where v i,j Let t' be the transport speed from node i to node j; i Let be the dwell time at node i;

[0078] The calculation strategy for the transportation risk R is as follows: Where r i,j Let be the transportation risk coefficient from node i to node j; it is determined based on road conditions, weather, and traffic congestion.

[0079] The improved ant colony algorithm in path planning module 4 introduces an adaptive pheromone update strategy and a local search operator on the basis of the ant colony algorithm. The adaptive pheromone update strategy is specifically: dynamically adjusting the pheromone evaporation coefficient and increment according to the quality of the path; the local search operator is used to optimize the initial solution obtained by the ant colony algorithm.

[0080] The improved pheromone update formula for the ant colony algorithm is as follows: Where τ i,j (t+1) represents the pheromone concentration from node i to node j at time t+1; ρ(t) represents the pheromone volatility coefficient at time t; τ i,j (t) represents the pheromone concentration from node i to node j at time t; m represents the number of ants in the colony. Let be the pheromone increment of the k-th ant on the path from node i to node j at time t;

[0081] The adaptive adjustment formula for the pheromone volatile coefficient ρ(t) is as follows: Where ρ min ρ is the minimum pheromone evaporation coefficient. max γ is the maximum pheromone evaporation coefficient; γ is the adjustment coefficient; iter is the current iteration number; iter max This represents the maximum number of iterations.

[0082] The pheromone increment The calculation formula is: Where Q represents the total amount of pheromones; F k The value of the multi-objective optimization function for the path taken by the k-th ant.

[0083] During the algorithm initialization phase, the ant colony size is set to 50-100 (dynamically adjusted based on the number of nodes; the more nodes, the larger the ant colony size can be), the initial pheromone concentration is set to 0.1-0.5, and the maximum number of iterations is set to 100-200. During path construction, ants select the next node based on a heuristic function. The larger the heuristic function value, the better the path is in the multi-objective comprehensive evaluation, and the higher the probability that the ant will choose that path.

[0084] The adaptive strategy employed during the pheromone update phase effectively prevents the algorithm from getting trapped in local optima. In the early stages of iteration, the pheromone evaporation coefficient ρ(t) takes a relatively large value (close to ρ). max (e.g., 0.8), which accelerates pheromone evaporation, allowing ants to explore paths more extensively; as the number of iterations increases, ρ(t) gradually decreases (approaching ρ). min To reduce pheromone evaporation, the γ value is typically set to 1-3 to control the decay rate of ρ(t). Simultaneously, a 2-opt local search operator is introduced to optimize the initial solution. For example, when a path intersects or detours, two nodes in the path are swapped (e.g., changing the path "node 1→node 3→node 2→node 4" to "node 1→node 2→node 3→node 4"), and the multi-objective optimization function value is recalculated. If the optimized function value is smaller, the path is updated to further improve the quality of the solution.

[0085] The inventory optimization module 5 uses an improved economic order quantity model, namely the EOQ model, to calculate the optimal order quantity and order cycle based on demand forecasts, inventory data, and supplier data.

[0086] The improved EOQ model in inventory optimization module 5 introduces a safety stock factor and a supply delay coefficient based on the EOQ model. The optimal order quantity calculation strategy of the improved EOQ model is as follows: Q * The optimal order quantity is σ; D is the annual demand; S is the cost per order; H is the annual holding cost per unit of inventory; σ is the minimum order quantity. D The standard deviation of demand; μ D The mean demand is represented by z; the safety stock factor is determined based on the service level; and L is the supply delay time.

[0087] The optimal order cycle calculation strategy of the improved EOQ model is as follows: Where T * To achieve the optimal ordering cycle;

[0088] Uncertainties in the actual supply chain are taken into account. Specifically, demand fluctuations are represented by the demand standard deviation σ. D and demand mean μ DTo quantify, σ D / μ D The larger the ratio, the more volatile the demand; the model will adjust accordingly by increasing the ratio. The numerical value of this item is used to increase the base order quantity to cope with demand fluctuations. Supply delays are addressed through supply delay time L and safety stock factor z. The z value is determined based on service level requirements; for example, z is 1.64 when the service level requirement is 95%, and 2.33 when the service level requirement is 99%. The safety stock portion... It can effectively reduce the risk of stockouts caused by supply delays.

[0089] For example, a fast-moving consumer goods (FMCG) company has an annual demand of D = 10,000 units, an ordering cost of S = 500 yuan, an annual holding cost per unit of inventory H = 20 yuan, and a demand standard deviation σ. D =500 units, average demand μ D =800 units, service level requirement 95% (z = 1.64), supply delay time L = 10 days. Substituting into the improved EOQ model calculation formula, we can obtain: basic order quantity The optimal order quantity is Q. The safety stock is approximately 1.64 × 500 × 10 ≈ 1.64 × 500 × 3.16 ≈ 2598.4 units. * ≈1154.7 + 2598.4 = 3753.1 units (rounded to 3753 units), optimal ordering cycle. This provides companies with accurate information for formulating order plans.

[0090] The dynamic monitoring module 6 collects operational data from each link of the logistics supply chain in real time, monitors the transportation process, inventory level and order execution, and feeds back abnormal information to the decision support module 7.

[0091] The system collects operational data from all stages of the logistics supply chain in real time through Internet of Things (IoT) devices. For example, GPS positioning devices acquire the real-time location and speed of transport vehicles, a warehouse management system (WMS) collects inventory quantities and inventory turnover rates, and an order management system (OMS) tracks order execution progress. When analyzing the collected data in real time, a threshold-based method is used to identify anomalies. For instance, the threshold for vehicle deviation (D) is set to 0.1. When D > 0.1, it is determined that the vehicle has deviated from the optimal route, and the system immediately sends a route adjustment reminder to the driver and feeds back the anomaly information to the decision support module. The inventory level threshold is set based on safety stock. When the inventory quantity falls below the safety stock, an inventory warning is triggered, reminding managers to replenish stock in a timely manner. The order execution delay threshold is set to 24 hours. When the order execution delay exceeds 24 hours, the system automatically analyzes the cause of the delay (such as transportation delays, inventory shortages, etc.) and generates corresponding solutions.

[0092] The decision support module 7 generates decision suggestions based on demand forecast results, optimal transportation routes, optimal inventory plans, and abnormal information, providing support for logistics supply chain management.

[0093] Based on the outputs of each module and combined with the overall operational goals of the logistics supply chain (such as reducing costs, improving efficiency, and enhancing customer satisfaction), a multi-attribute decision-making method is used to generate decision recommendations. For example, when demand forecasting shows that demand for a certain type of product will increase by 20% in the next month and the current inventory level is below the safety stock, the decision support module will suggest increasing the order quantity for that type of product and adjusting the transportation plan according to the optimal transportation route output by the route planning module to ensure timely replenishment of products. When the dynamic monitoring module reports that the transportation risk of a certain transportation route has increased significantly due to severe weather, the decision support module will suggest adjusting the transportation route, selecting an alternative route with lower transportation risk, and recalculating transportation costs and time to ensure the stable operation of the overall supply chain.

[0094] Example 2

[0095] Please see Figure 2 A dynamic logistics supply chain management method based on big data, comprising the following steps:

[0096] S1: Data collection, which collects data from all aspects of the logistics supply chain through the big data collection module 1, including customer demand data, transportation data, inventory data, supplier data and market environment data;

[0097] S2: Data preprocessing. Data preprocessing module 2 cleans the collected raw data, removing outliers and missing values; integrates the cleaned data, combining data from different sources into a unified data warehouse; transforms the integrated data into a format suitable for model analysis; and reduces the dimensionality and quantity of the transformed data to obtain a standardized dataset.

[0098] S3: Demand Forecasting. Demand Forecasting Module 3 uses an improved LSTM demand forecasting model to forecast customer demand data in a standardized dataset.

[0099] S31. Normalize the customer demand data and map the data to the [0,1] range;

[0100] S32. Divide the normalized customer demand data into a training set and a test set according to the time series;

[0101] S33. Train the improved LSTM demand prediction model using the training set, and adjust the model parameters through the backpropagation algorithm to minimize the prediction error of the model;

[0102] S34. Use the test set to validate the trained model. If the model's prediction accuracy meets the preset requirements, output the predicted value of customer demand for a future period of time; otherwise, readjust the model parameters and train it.

[0103] The normalization process uses the Min-Max normalization method, specifically: Where x′ represents the normalized data, x represents the original data, and x... min x is the minimum value in the original data. max This represents the maximum value in the original data.

[0104] S4: Route planning. Route planning module 4 constructs a multi-objective route planning model based on demand forecasts, transportation data, and real-time traffic data, and uses an improved ant colony algorithm to solve the multi-objective route planning model.

[0105] S41. Initialize the parameters of the ant colony algorithm, including the number of ants, the initial pheromone concentration, and the maximum number of iterations;

[0106] S42. Each ant selects the next node based on the pheromone concentration between nodes and a heuristic function, thus constructing a transportation path;

[0107] The heuristic function is as follows: Where η i,j C represents the heuristic function value from node i to node j. i,j Let T′ be the unit transportation cost from node i to node j. i,jR is the unit transport time from node i to node j. i,j The unit transportation risk from node i to node j;

[0108] The local search operator adopts a 2-opt swapping strategy, swapping two nodes in the path and recalculating the multi-objective optimization function value of the path. If the swapped path is better, the new path is retained; otherwise, the original path is retained.

[0109] S43. Calculate the multi-objective optimization function value of the path taken by each ant, and update the pheromone concentration between nodes according to the adaptive pheromone update strategy;

[0110] S44. Introduce local search operators to optimize the initial solution and improve the quality of the solution;

[0111] S45. When the number of iterations reaches the maximum number of iterations or the quality of the solution meets the preset requirements, output the optimal transportation path;

[0112] S5: Inventory Optimization. The Inventory Optimization Module 5 uses an improved EOQ model to calculate the optimal order quantity and order cycle based on demand forecasts, inventory data, and supplier data.

[0113] S51. Obtain parameters such as annual demand, cost per order, annual holding cost per unit of inventory, standard deviation of demand, mean demand, service level, and supply delay time.

[0114] S52 substitutes the parameters obtained in S51 into the optimal order quantity calculation strategy and optimal order cycle calculation strategy of the improved EOQ model to obtain the optimal order quantity and order cycle.

[0115] S6: Dynamic monitoring. The dynamic monitoring module 6 collects operational data from all aspects of the logistics supply chain in real time, including the location, speed, inventory level, and order execution progress of transport vehicles. It analyzes the collected operational data in real time to determine whether the transportation process, inventory level, and order execution are normal. If any abnormalities are found, such as transport vehicles deviating from the optimal route, inventory levels falling below safety stock, or order execution delays, the abnormal information is fed back to the decision support module 7.

[0116] The judgment of abnormal situations adopts the threshold judgment method. Threshold ranges are set in advance for each operating indicator. If the operating indicator exceeds the threshold range, it is judged as an abnormal situation.

[0117] The operational metrics include vehicle deviation, inventory levels, and order fulfillment delay time.

[0118] The formula for calculating the deviation of the transport vehicle is: Where D′ is the deviation of the transport vehicle; m′ is the number of monitoring time intervals; d tD'0 represents the distance the transport vehicle deviates from the optimal path during the t-th time interval; D'0 represents the total distance along the optimal path.

[0119] S7: Decision Support. Decision Support Module 7 generates decision suggestions based on demand forecast results, optimal transportation routes, optimal inventory plans, and abnormal information, combined with the overall operational goals of the logistics supply chain. These suggestions include adjusting transportation plans, optimizing inventory structure, and negotiating with suppliers to adjust delivery times. The decision suggestions are then presented to logistics supply chain managers in a visual manner to support their decision-making.

[0120] The visualization uses charts, including line charts, bar charts, pie charts, and maps. Line charts are used to show demand forecast trends and inventory level changes, bar charts are used to show cost and time comparisons for different transportation routes, pie charts are used to show inventory structure distribution, and maps are used to show the real-time location of transportation vehicles and optimal routes.

[0121] Example 3

[0122] To facilitate understanding of the technical solution of the present invention, this embodiment is described in detail with reference to a specific application scenario:

[0123] First, the system hardware and software are deployed. On the hardware side, a big data server (such as a Hadoop cluster, including 3 master nodes and 10 slave nodes) is configured for data storage and computation. GPS positioning devices (installed on transport vehicles) and RFID identification devices (installed at warehouse entrances and exits) are deployed for data collection. Visual terminals (such as computers and mobile phones for managers) are configured for display and operation. On the software side, the core modules of the system are developed using Java, structured data is stored using a MySQL database, unstructured data is stored using an HBase database, an improved LSTM demand forecasting model is implemented using the TensorFlow framework, an improved ant colony algorithm is implemented using Python, and data visualization is achieved using ECharts.

[0124] Big Data Acquisition Module 1 collects data in the following ways:

[0125] Customer demand data: obtained through the e-commerce platform's order management system, collecting information such as customer order quantity, order time, product type, and shipping address in real time, with a data collection frequency of 1 minute / time;

[0126] Transportation data: Real-time location, speed, and direction of travel of transport vehicles are collected through GPS positioning devices. Transportation routes, transportation costs, and transportation time are obtained through the transportation management system. The data collection frequency is 30 seconds per time.

[0127] Inventory data: Information such as inventory quantity, inventory location, inventory turnover rate, product inbound time, and outbound time are collected through the warehouse management system and RFID identification equipment. The data collection frequency is 5 minutes / time.

[0128] Supplier data: Information such as supplier supply capacity, delivery time, supply cost, and product quality pass rate is obtained through the supplier management system. The data collection frequency is once a day.

[0129] Market environment data: Information such as market prices, market demand trends, and competitors' product sales and prices are obtained from e-commerce platforms and industry report websites through web crawling technology. The data collection frequency is once a day.

[0130] Data preprocessing module 2 processes the collected raw data:

[0131] Data cleaning: Box plots are used to identify outliers (such as data whose transportation time exceeds the normal range by more than 3 times), mean imputation is used to fill missing values ​​(such as filling missing inventory quantity data with the average inventory quantity of the product in the previous 7 days), and hash tables are used to remove duplicate data.

[0132] Data integration: Integrate data from different systems (such as order management system, transportation management system, and warehouse management system) into the HBase data warehouse using unified identifiers (such as product ID and order ID) to eliminate data redundancy and conflicts;

[0133] Data transformation: Convert text-type data (such as product category names) into numerical data (such as using 1 to represent food and 2 to represent clothing), and use the Min-Max normalization method to normalize the numerical data (such as order quantity and transportation cost);

[0134] Data reduction: Principal component analysis (PCA) is used to reduce the dimensionality of market environment data, reducing the original 20 dimensions of market data to 5 principal components, retaining more than 90% of the information and reducing the amount of data processing.

[0135] The implementation steps of demand forecasting module 3 are as follows:

[0136] Data preparation: Extract customer demand data from the preprocessed standardized dataset, classify them according to product type, and form an independent time series data for each product type with a time granularity of 1 day;

[0137] Model training: The time series data of each product category are divided into training set and test set in a 7:3 ratio. The improved LSTM demand prediction model is trained using the training set. The model has 2 hidden layers, 64 neurons per layer, a learning rate of 0.001, and 100 iterations. The root mean square error (RMSE) of the model is minimized using the Adam optimizer.

[0138] Model validation: Use a test set to validate the trained model and calculate the required prediction accuracy. If the accuracy reaches 90% or more, the model is qualified. If the accuracy does not meet the requirements, adjust the model parameters (such as increasing the number of hidden layer neurons or adjusting the learning rate) and retrain until the accuracy meets the requirements.

[0139] Demand Forecasting: Using a qualified model, forecast customer demand for the next 30 days, output the daily demand forecast value for each product category for the next 30 days, store the forecast results in the MySQL database, and send them to the inventory optimization module 5 and the decision support module 7.

[0140] The implementation steps of path planning module 4 are as follows:

[0141] Data Input: Receives demand forecast values ​​(determining the quantity and destination of transported goods), transportation data (including vehicle information, distance between nodes, unit distance transportation cost, transportation speed, and transportation risk coefficient) and real-time traffic data (such as congested road information) output from demand forecast module 3; Model Construction: Constructs a multi-objective path planning model based on the input data, and determines the weight coefficients of transportation cost, transportation time, and transportation risk. For example, for fresh produce transportation, ω1 = 0.3, ω2 = 0.4, and ω3 = 0.2 are set.

[0142] Algorithm Solution: An improved ant colony algorithm is used to solve the multi-objective path planning model. The initial ant colony size is 80, the initial pheromone concentration is 0.3, the maximum number of iterations is 150, and the pheromone evaporation coefficient ranges from ρ. max =0.8, ρ min =0.2, adjustment coefficient γ=2. Each ant selects the next node based on the pheromone concentration between nodes and the heuristic function value, and constructs the initial transportation path;

[0143] Path optimization: The initial path is optimized by applying the 2-opt local search operator. For example, when an intersection is detected in the path, the intersection is eliminated by swapping the order of the two nodes, the multi-objective optimization function value of the path is recalculated, and the better path is retained.

[0144] Output results: When the algorithm reaches the maximum number of iterations or the change in the path optimization function value is less than 0.01 after 10 consecutive iterations, the optimal transportation route is output, including information such as transportation route, nodes along the way, estimated transportation time, transportation cost and transportation risk. The results are then sent to the dynamic monitoring module 6 and the decision support module 7.

[0145] The implementation steps for Inventory Optimization Module 5 are as follows:

[0146] Parameter acquisition: Obtain the annual demand D for each product type from the demand forecast module 3 (calculated based on the demand forecast value for the next 30 days), obtain the order cost S and supply delay time L from the supplier data, obtain the annual holding cost per unit of inventory H, demand standard deviation σD and demand mean μD from the inventory data, and determine the safety stock factor z according to the company's service level requirements (e.g., z = 1.64 for a service level of 95%).

[0147] Calculate the optimal order quantity and lead time: Substitute the above parameters into the improved EOQ model calculation formula to calculate the optimal order quantity Q and optimal lead time T for each product type.

[0148] Generate inventory plan: Based on the optimal order quantity and order cycle, and combined with the current inventory level, generate order plans for each product type, including suggested order time, order quantity, supplier selection and other information, and send the inventory plan to the decision support module 7 and the dynamic monitoring module 6.

[0149] The implementation steps of dynamic monitoring module 6 are as follows:

[0150] Real-time data collection: Information such as the location, speed, and fuel consumption of transport vehicles, warehouse inventory and inbound / outbound records, and order execution progress, delivery status, and receipt status are collected in real time through IoT devices. The data collection frequency is dynamically adjusted according to the data type (e.g., transport vehicle location every 30 seconds, inventory quantity every 5 minutes).

[0151] Real-time data analysis: Analyze the collected real-time data, calculate operational indicators such as vehicle deviation D, inventory level, and order execution delay time, and compare them with preset thresholds;

[0152] Anomaly detection and feedback: When the operating indicators exceed the threshold range, it is determined to be an anomaly. The system automatically generates an anomaly information report, including the anomaly type, occurrence time, and scope of impact, and feeds the report back to the decision support module in real time. At the same time, relevant management personnel are notified via SMS, email, and other means.

[0153] The implementation steps for Decision Support Module 7 are as follows:

[0154] Data integration: Receive the demand forecast results from the demand forecasting module 3, the optimal transportation route from the route planning module 4, the inventory plan from the inventory optimization module 5, and the abnormal information from the dynamic monitoring module 6, and integrate and perform correlation analysis on these data.

[0155] Decision recommendation generation: Based on the integrated data analysis results and combined with the company's operational goals (such as reducing costs, improving efficiency, and enhancing customer satisfaction), decision recommendations are generated using the Analytic Hierarchy Process (AHP). For example, when demand forecasts indicate a surge in demand for a certain product and current inventory is insufficient, it is recommended to immediately initiate an emergency ordering process and adjust transportation routes to prioritize the delivery of this product.

[0156] Visualization: Information such as decision-making suggestions, demand forecast trends, inventory level changes, and transportation routes are displayed on the terminal devices of managers through visual charts (line charts, bar charts, maps, etc.), supporting managers to perform interactive operations (such as viewing detailed data, adjusting parameters, etc.);

[0157] Decision execution tracking: Track the decision recommendations adopted by managers, monitor the implementation and effects of decisions in real time, such as cost changes after transportation route adjustments and inventory level changes after ordering plan execution, and feed the tracking results back to managers.

[0158] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A dynamic logistics supply chain management system based on big data, comprising a big data acquisition module (1), a data preprocessing module (2), a demand forecasting module (3), a route planning module (4), an inventory optimization module (5), a dynamic monitoring module (6), and a decision support module (7), characterized in that: The big data acquisition module (1) is used to collect data from all aspects of the logistics supply chain, including customer demand data, transportation data, inventory data, supplier data and market environment data. The data preprocessing module (2) is used to clean, integrate, transform and reduce the collected raw data to obtain a standardized dataset. The demand forecasting module (3) uses an improved long short-term memory network, i.e., LSTM demand forecasting model, to predict the preprocessed customer demand data and output the predicted customer demand value for a future period of time. The path planning module (4) is based on the improved ant colony algorithm, combined with transportation data and real-time traffic data, to construct a multi-objective path planning model and output the optimal transportation path. The inventory optimization module (5) uses the improved economic order quantity model, namely the EOQ model, to calculate the optimal order quantity and order cycle based on demand forecasts, inventory data and supplier data. The dynamic monitoring module (6) collects operational data of each link in the logistics supply chain in real time, monitors the transportation process, inventory level and order execution, and feeds back abnormal information to the decision support module (7). The decision support module (7) generates decision suggestions based on demand forecast results, optimal transportation routes, optimal inventory plans and abnormal information, providing support for logistics supply chain management.

2. The big data-based dynamic logistics supply chain management system according to claim 1, characterized in that, The output of the big data acquisition module (1) is electrically connected to the input of the data preprocessing module (2), and the output of the data preprocessing module (2) is electrically connected to the input of the demand forecasting module (3). The improved LSTM demand forecasting model in the demand forecasting module (3) introduces an attention mechanism and a time decay factor on the basis of the LSTM model. The attention mechanism is used to assign different weights to the input data at different time steps, and the time decay factor is used to adjust the degree of influence of historical data. The formula for calculating the output layer of the improved LSTM demand forecasting model is as follows: Where y t W represents the demand forecast at time t. o b is the output layer weight matrix; o This is the output bias vector; The cell state after adjustment by attention mechanisms and time decay factors; The weight calculation formula for the attention mechanism is as follows: Where α t,i Let e ​​be the attention weight for the input data at time t to the i-th time step; t,i A score for attention; v a W a U a b is the parameter matrix of the attention mechanism; a h is the bias vector for the attention mechanism. i h represents the hidden state at the i-th time step. t Let be the hidden state at time t; T represents the total number of time steps; The formula for calculating the time decay factor is: λ t,i =exp(-β·(ti)), where λ t,i Let be the time decay factor of the input data at time t for the i-th time step; β is the attenuation coefficient, and its value ranges from 0.01 to 0.

1.

3. The big data-based dynamic logistics supply chain management system according to claim 2, characterized in that, The multi-objective route planning model takes the minimum transportation cost, the shortest transportation time, and the minimum transportation risk as its objective functions. The expression of the objective function is: minF=(ω1·C,ω2·T′,ω3·R), where F is the multi-objective optimization function; ω1, ω2, and ω3 are the weight coefficients of transportation cost, transportation time, and transportation risk, respectively, and ω1+ω2+ω3=1; C is the transportation cost; T′ is the transportation time; and R is the transportation risk. The calculation strategy for the transportation cost C is as follows: Where n is the total number of nodes; d i,j Let be the distance from node i to node j; c is the transportation cost per unit distance; x i,j It is a 0-1 variable, and x i,j =1 indicates that there is a transportation path from node i to node j, x i,j =0 indicates no transport path; f i y represents the fixed cost of node i; i It is a 0-1 variable, and y i =1 indicates that node i is a stop point, y i =0 indicates no selection; The calculation strategy for the transportation time T′ is as follows: Where v i,j Let t' be the transport speed from node i to node j; i Let be the dwell time at node i; The calculation strategy for the transportation risk R is as follows: Where r i,j Let be the transportation risk coefficient from node i to node j.

4. The big data-based dynamic logistics supply chain management system according to claim 3, characterized in that, The improved ant colony algorithm in the path planning module (4) introduces an adaptive pheromone update strategy and a local search operator on the basis of the ant colony algorithm. The adaptive pheromone update strategy is specifically: dynamically adjusting the pheromone evaporation coefficient and increment according to the quality of the path. The local search operator is used to optimize the initial solution obtained by the ant colony algorithm; The improved pheromone update formula for the ant colony algorithm is as follows: Where τ i,j (t+1) represents the pheromone concentration from node i to node j at time t+1; ρ(t) represents the pheromone volatility coefficient at time t; τ i,j (t) represents the pheromone concentration from node i to node j at time t; m represents the number of ants in the colony. Let be the pheromone increment of the k-th ant on the path from node i to node j at time t; The adaptive adjustment formula for the pheromone volatile coefficient ρ(t) is as follows: Where ρ min ρ is the minimum pheromone evaporation coefficient. max γ is the maximum pheromone evaporation coefficient; γ is the adjustment coefficient; iter is the current iteration number; iter max This represents the maximum number of iterations. The pheromone increment The calculation formula is: Where Q represents the total amount of pheromones; F k The value of the multi-objective optimization function for the path taken by the k-th ant.

5. The big data-based dynamic logistics supply chain management system according to claim 4, characterized in that, The improved EOQ model in the inventory optimization module (5) introduces a safety stock factor and a supply delay coefficient based on the EOQ model. The optimal order quantity calculation strategy of the improved EOQ model is as follows: Q * The optimal order quantity is σ; D is the annual demand; S is the cost per order; H is the annual holding cost per unit of inventory; σ is the minimum order quantity. D The standard deviation of demand; μ D Z represents the mean demand; z represents the safety stock factor; and L represents the supply delay time. The optimal order cycle calculation strategy for the improved EOQ model is as follows: Where T * This is the optimal ordering cycle.

6. A big data-based dynamic management method for logistics supply chains, applied to the big data-based dynamic management system for logistics supply chains as described in any one of claims 1-5, characterized in that, The management method includes the following steps: S1: Data collection, through the big data collection module (1) to collect data from all aspects of the logistics supply chain, including customer demand data, transportation data, inventory data, supplier data and market environment data; S2: Data preprocessing, the data preprocessing module (2) cleans the collected raw data, removes outliers and missing values; integrates the cleaned data, integrates data from different sources into a unified data warehouse; and transforms the integrated data into a format suitable for model analysis. S3: Demand forecasting, the demand forecasting module (3) uses an improved LSTM demand forecasting model to forecast customer demand data in a standardized dataset; S31. Normalize the customer demand data and map the data to the [0,1] range; S32. Divide the normalized customer demand data into a training set and a test set according to the time series; S33. Train the improved LSTM demand prediction model using the training set, and adjust the model parameters through the backpropagation algorithm to minimize the prediction error of the model; S34. Use the test set to validate the trained model. If the model's prediction accuracy meets the preset requirements, output the predicted value of customer demand for a future period of time; otherwise, readjust the model parameters and train it. S4: Route planning, the route planning module (4) constructs a multi-objective route planning model based on demand forecasts, transportation data and real-time traffic data, and uses an improved ant colony algorithm to solve the multi-objective route planning model; S41. Initialize the parameters of the ant colony algorithm, including the number of ants, the initial pheromone concentration, and the maximum number of iterations; S42. Each ant selects the next node based on the pheromone concentration between nodes and a heuristic function, thus constructing a transportation path; S43. Calculate the multi-objective optimization function value of the path taken by each ant, and update the pheromone concentration between nodes according to the adaptive pheromone update strategy; S44. Introduce a local search operator to optimize the initial solution; S45. When the number of iterations reaches the maximum number of iterations or the quality of the solution meets the preset requirements, output the optimal transportation path; S5: Inventory optimization, the inventory optimization module (5) calculates the optimal order quantity and order cycle based on demand forecasts, inventory data and supplier data using an improved EOQ model; S51. Obtain parameters such as annual demand, cost per order, annual holding cost per unit of inventory, standard deviation of demand, mean demand, service level, and supply delay time. S52 substitutes the parameters obtained in S51 into the optimal order quantity calculation strategy and optimal order cycle calculation strategy of the improved EOQ model to obtain the optimal order quantity and order cycle. S6: Dynamic monitoring. The dynamic monitoring module (6) collects real-time operational data of each link in the logistics supply chain, including the location, speed, inventory level, and order execution progress of transport vehicles; it performs real-time analysis on the collected operational data to determine whether the transportation process, inventory level, and order execution are normal; if an abnormal situation is found, the abnormal information is fed back to the decision support module (7). S7: Decision support. The decision support module (7) generates decision suggestions based on demand forecast results, optimal transportation routes, optimal inventory plans and abnormal information, combined with the overall operational goals of the logistics supply chain; and displays the decision suggestions to logistics supply chain managers in a visual manner.

7. The big data-based dynamic management method for logistics supply chains according to claim 6, characterized in that, In S3, the normalization process uses the Min-Max normalization method, specifically: Where x′ represents the normalized data, x represents the original data, and x... min x is the minimum value in the original data. max This represents the maximum value in the original data.

8. The big data-based dynamic management method for logistics supply chains according to claim 7, characterized in that, In S4, the heuristic function is specifically: Where η i,j C represents the heuristic function value from node i to node j. i,j Let T′ be the unit transportation cost from node i to node j. i,j R is the unit transport time from node i to node j. i,j The unit transportation risk from node i to node j; The local search operator employs a 2-opt swapping strategy, which swaps two nodes in the path and recalculates the multi-objective optimization function value of the path. If the swapped path is better, the new path is retained; otherwise, the original path is retained.

9. The method for dynamic management of logistics supply chain based on big data according to claim 6, characterized in that, In S6, the abnormal situation is judged by a threshold judgment method. A threshold range is set in advance for each operating indicator. If the operating indicator exceeds the threshold range, it is judged as an abnormal situation. The operational metrics include vehicle deviation, inventory levels, and order fulfillment delay time. The formula for calculating the deviation of the transport vehicle is as follows: Where D′ is the deviation of the transport vehicle; m′ is the number of monitoring time intervals; d t D'0 represents the distance the transport vehicle deviates from the optimal path during the t-th time interval; D'0 represents the total distance of the optimal path.

10. The method for dynamic management of logistics supply chain based on big data according to claim 6, characterized in that, In S7, the visualization is presented in the form of charts, including line charts, bar charts, pie charts, and maps. Line charts are used to show demand forecast trends and inventory level change trends, bar charts are used to show the cost and time comparison of different transportation routes, pie charts are used to show the inventory structure distribution, and maps are used to show the real-time location and optimal route of transportation vehicles.

Images