A data-driven intelligent logistics supply chain digital management system

Abstract

This invention relates to a data-driven intelligent logistics supply chain digital management system. The system includes a demand forecasting and analysis unit, an inventory optimization unit, a supplier management unit, a logistics route optimization unit, a system integration and feedback unit, and a real-time logistics marketing unit. By analyzing historical orders and market data through recurrent neural networks and long short-term memory networks, accurate forecasting of short-term and medium-to-long-term demand is achieved. The inventory optimization unit combines a Bayesian dynamic linear model and a convolutional neural network to achieve dynamic inventory management. The supplier management unit optimizes supplier selection using a multilayer perceptron and fuzzy clustering algorithm. The logistics route optimization unit achieves optimal scheduling of logistics routes based on genetic algorithms and an adaptive large-scale neighborhood search algorithm (ALNS). By integrating multiple functional units, this invention effectively improves the operational efficiency and response speed of the logistics supply chain.

Description

Technical Field

[0001] This invention relates to the field of supply chain management, and more particularly to a smart logistics supply chain digital management system based on data analysis. Background Technology

[0002] Currently, the main challenge facing supply chain management systems lies in improving the collaborative efficiency of each link through intelligent means, especially in dealing with demand fluctuations, inventory management, and supplier relationship management. Traditional logistics supply chain management systems rely on manual experience and cannot accurately analyze market data and historical order data, resulting in inaccurate forecasts and difficulty in adapting to dynamic changes in market demand. Existing inventory management systems are difficult to adjust in real time and cannot effectively cope with dynamic changes in inventory, which may lead to excess or shortage of inventory and increase inventory costs. Supplier selection relies on historical performance and lacks analysis of real-time supplier data, resulting in an inability to respond to demand in a timely manner during periods of high demand fluctuations, affecting the overall efficiency of logistics. Summary of the Invention

[0003] This invention provides a smart logistics supply chain digital management system based on data analysis to solve the above-mentioned technical problems.

[0004] This invention provides a data-driven intelligent logistics supply chain digital management system, the system comprising:

[0005] The demand forecasting and analysis unit analyzes historical order and market fluctuation data through recurrent neural networks and long short-term memory networks to predict short-term and medium- to long-term demand and output demand forecast data.

[0006] The inventory optimization unit, combining a Bayesian dynamic linear model and a convolutional neural network algorithm, adjusts the inventory level based on the demand forecast data output by the demand forecasting and analysis unit.

[0007] Supplier Management Unit: Based on multilayer perceptron and fuzzy clustering algorithm, it receives inventory adjustment data from the inventory optimization unit and selects, classifies and sorts suppliers according to the supplier data; this unit also formulates supplier delivery plans and adjusts order allocation strategies by combining the supplier's delivery capabilities and historical performance.

[0008] The logistics route optimization unit uses a genetic algorithm and an adaptive large-scale neighborhood search algorithm to plan logistics routes based on the supplier selection results of the supplier management unit and real-time logistics data.

[0009] The system integration and feedback unit receives data from the demand forecasting and analysis unit, the inventory optimization unit, the supplier management unit, and the logistics route optimization unit through a global data management and feedback mechanism, and integrates and synchronizes the operation of each unit.

[0010] The real-time logistics marketing unit receives feedback data from the system integration and feedback unit and adjusts logistics marketing strategies based on logistics data and market trends.

[0011] Furthermore, the demand forecasting and analysis unit includes:

[0012] Short-term forecasting module: Generates short-term demand forecasts through a recurrent neural network and transmits the forecasts to the inventory optimization unit to adjust inventory levels;

[0013] Medium- and long-term forecasting module: It uses a long short-term memory network to generate medium- and long-term demand forecasts, and combines the output of the long short-term memory network with the short-term forecast results of the recurrent neural network algorithm to transmit to the inventory optimization unit and the logistics route optimization unit to formulate long-term inventory and logistics scheduling plans.

[0014] Attention mechanism module: Assigns different weights to the prediction data output by the long short-term memory network, and transmits the weighted prediction data to the supplier management unit and the logistics route optimization unit;

[0015] Feedback module: Collects operational data from the inventory optimization unit, the logistics path unit, and the supplier management unit, and transmits feedback information to the recurrent neural network and the long short-term memory network to adjust model parameters according to actual operation.

[0016] Furthermore, the long short-term memory network in the medium- and long-term prediction module of the demand prediction and analysis unit includes an input gate, a forget gate, and an output gate. The input gate is used to select the current input information, the forget gate is used to adjust the weight of historical data on the current prediction, and the output gate is used to generate the prediction result.

[0017] Furthermore, the inventory optimization unit includes:

[0018] The data acquisition module is used to acquire data from the demand forecasting and analysis unit, supplier management unit, logistics route optimization unit, and inventory optimization unit. The data includes historical order data, real-time order volume, market fluctuation data, supplier delivery information, and logistics status data. After standardization processing, the data is transmitted to the inventory forecasting module and inventory data module to provide a basis for inventory forecasting and adjustment.

[0019] Inventory forecasting module: combines a Bayesian dynamic linear model and a convolutional neural network for inventory forecasting; the Bayesian dynamic linear model is used for short-term inventory forecasting, and the convolutional neural network is used to analyze long-term inventory trends; the forecast results are transmitted to the inventory adjustment module to provide a basis for short-term and long-term inventory control.

[0020] Inventory Adjustment Module: Based on the forecast results of the inventory forecasting module, dynamically adjust the inventory level;

[0021] Feedback processing module: Collects operational feedback information from the supplier management unit and the logistics route optimization unit, and transmits the operational feedback information to the inventory forecasting module and the inventory adjustment module.

[0022] Furthermore, the convolutional neural network in the inventory optimization unit includes multiple convolutional layers and pooling layers. The convolutional layers are used to process local patterns in the inventory data, and the pooling layers are used to reduce data dimensionality and extract long-term trend features of the inventory data.

[0023] Furthermore, the supplier management unit includes:

[0024] Supplier data acquisition module: used to acquire relevant supplier data from various units of the supply chain, including supplier delivery time, order completion rate, inventory demand data and logistics status information; after the acquired data is standardized, it is transmitted to the supplier analysis module and the supplier classification and ranking module.

[0025] Supplier Analysis Module: Based on the multilayer perceptron algorithm, it analyzes the supplier's historical delivery records, response speed, and delivery stability, and generates supplier priorities; based on the analysis results, it adjusts the supplier delivery plan and transmits the supplier priority ranking results to the supplier classification and ranking module;

[0026] Supplier classification and ranking module: Uses fuzzy clustering algorithm to classify and rank suppliers; determines the supplier classification results based on supplier response speed, delivery stability and inventory requirements and transmits them to the inventory optimization unit and logistics route optimization unit;

[0027] Supplier feedback processing module: used to collect and process operational data from other units in the supply chain, including supplier delivery time, logistics execution status and inventory turnover rate; and to transmit the feedback data to the supplier analysis module.

[0028] Furthermore, the logistics route optimization unit includes:

[0029] Logistics data acquisition module: used to acquire logistics-related data, including inventory status, demand forecast information, traffic conditions and transport vehicle status information; the acquired data is processed and then transmitted to the route planning module and resource scheduling module;

[0030] Path planning module: Generates optimal path solutions based on genetic algorithm and Dijkstra's algorithm; genetic algorithm is used to generate initial path solution set, Dijkstra's algorithm calculates the shortest path solution based on the initial solution and passes the solution to resource scheduling module;

[0031] Resource scheduling module: Based on the route plan generated by the route planning module, it uses an adaptive large-scale neighborhood search algorithm to schedule logistics resources;

[0032] Route Feedback Module: Used to collect actual execution data of logistics routes, including transportation time, resource utilization rate and transportation delays; transmit feedback data to the route planning module and resource scheduling module to adjust route planning and resource scheduling schemes.

[0033] Furthermore, the initial path solution set of the genetic algorithm is set to 50 to 100.

[0034] Furthermore, the system integration and feedback unit includes:

[0035] Global Data Management Module: Used to acquire data from the demand forecasting and analysis unit, the inventory optimization unit, the supplier management unit, and the logistics route optimization unit; to standardize the acquired data and transmit it to the feedback optimization module and the intelligent collaboration module;

[0036] Feedback optimization module: Based on the data processed by the global data management module, and combined with inventory levels, logistics status, and the actual response of suppliers, adjust the operating parameters of the demand forecasting and analysis unit, the inventory optimization unit, and the logistics route optimization unit.

[0037] Intelligent Collaboration Module: By analyzing the data provided by the global data management module, it coordinates the operation of the demand forecasting and analysis unit, the inventory optimization unit, the supplier management unit, and the logistics route optimization unit;

[0038] Dynamic monitoring module: Identifies potential problems and provides early warning signals by monitoring logistics delays, inventory imbalances, and supplier responses.

[0039] Furthermore, the real-time logistics marketing unit includes:

[0040] Logistics data monitoring module: Used to monitor key data in the logistics process, including inventory levels, logistics route status, order completion status, and customer feedback information; the monitored data is transmitted to the market trend prediction module and the user behavior analysis module;

[0041] Market trend forecasting module: Based on historical sales data, market fluctuation data, and logistics data, it forecasts future market demand trends and transmits the forecast results to the user behavior analysis module and the marketing feedback module.

[0042] User behavior analysis module: By analyzing customer order history, purchasing behavior and logistics feedback data, it generates marketing plans for different customer groups; the generated marketing plans are then transmitted to the marketing feedback module and the inventory optimization unit.

[0043] Marketing Feedback Module: Used to collect execution data of marketing activities and customer feedback on logistics services; transmit feedback data to the market trend forecasting module and user behavior analysis module to adjust market forecasts and marketing strategies.

[0044] This invention provides a data-driven intelligent logistics supply chain digital management system. By integrating multiple functional units, it achieves comprehensive optimization of supply chain management and has the following beneficial effects:

[0045] 1. Improved accuracy in demand forecasting: By combining recurrent neural networks (RNNs) and long short-term memory networks (LSTMs), historical order and market fluctuation data can be effectively processed to generate accurate short-term and medium-to-long-term demand forecasts. This enables the system to respond quickly to demand fluctuations, ensuring supply and demand balance in the supply chain.

[0046] 2. Improved Inventory Management Efficiency: The inventory optimization unit employs a Bayesian dynamic linear model and a convolutional neural network to achieve dynamic inventory forecasting and adjustment. By combining short-term and long-term inventory forecasting, the system can flexibly adjust inventory levels when market demand changes, avoiding inventory backlog or shortages and reducing inventory holding costs.

[0047] 3. Supplier selection optimization: The supplier management unit can dynamically analyze the historical performance and real-time response data of suppliers through multilayer perceptron (MLP) and fuzzy clustering algorithms, thereby optimizing supplier selection, classification and ranking, ensuring the selection of the best suppliers during periods of high demand fluctuations, and improving the overall efficiency of the supply chain.

[0048] 4. Logistics Route Optimization: The logistics route optimization unit combines genetic algorithms, adaptive large-scale neighborhood search (ALNS) algorithms, and Dijkstra's algorithm to achieve optimal planning and resource scheduling of logistics routes. This ensures efficient utilization of logistics resources and reduces transportation delays and resource waste.

[0049] 5. Enhanced System Integration and Feedback Mechanism: Through global data management and feedback optimization via the system integration and feedback unit, real-time dynamic adjustments can be made to each unit of the supply chain. This module combines supplier response, inventory status, and logistics execution to ensure overall system optimization and efficient operation.

[0050] 6. Real-time Marketing Strategy Adjustment: Based on data analysis of logistics and market trends, the real-time logistics marketing unit can dynamically adjust marketing strategies to improve customer experience and order completion rates, while optimizing inventory marketing to ensure a close match between logistics and market demand. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 A schematic diagram of the system provided by the present invention;

[0053] Figure 2 This is a schematic diagram of the demand forecasting and analysis unit in this invention;

[0054] Figure 3 This is a schematic diagram of the inventory optimization unit in this invention;

[0055] Figure 4 This is a schematic diagram of the supplier management unit in this invention;

[0056] Figure 5 This is a schematic diagram of the logistics path optimization unit in this invention;

[0057] Figure 6 This is a schematic diagram of the system integration and feedback unit in this invention;

[0058] Figure 7 This is a schematic diagram of the structure of the real-time logistics marketing unit diagram in this invention. Detailed Implementation

[0059] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0060] Example 1: Please refer to Figure 1 This embodiment provides a data-driven intelligent logistics supply chain digital management system, the system comprising:

[0061] The demand forecasting and analysis unit analyzes historical order and market fluctuation data through recurrent neural networks and long short-term memory networks to predict short-term and medium- to long-term demand and output demand forecast data.

[0062] The inventory optimization unit, combining a Bayesian dynamic linear model and a convolutional neural network algorithm, adjusts the inventory level based on the demand forecast data output by the demand forecasting and analysis unit.

[0063] Supplier Management Unit: Based on multilayer perceptron and fuzzy clustering algorithm, it receives inventory adjustment data from the inventory optimization unit and selects, classifies and sorts suppliers according to the supplier data; this unit also formulates supplier delivery plans and adjusts order allocation strategies by combining the supplier's delivery capabilities and historical performance.

[0064] The logistics route optimization unit uses a genetic algorithm and an adaptive large-scale neighborhood search algorithm to plan logistics routes based on the supplier selection results of the supplier management unit and real-time logistics data.

[0065] The system integration and feedback unit receives data from the demand forecasting and analysis unit, the inventory optimization unit, the supplier management unit, and the logistics route optimization unit through a global data management and feedback mechanism, and integrates and synchronizes the operation of each unit.

[0066] The real-time logistics marketing unit receives feedback data from the system integration and feedback unit and adjusts logistics marketing strategies based on logistics data and market trends.

[0067] This embodiment utilizes recurrent neural networks (RNNs) and long short-term memory networks (LSTMs) to accurately predict short-term and medium-to-long-term demand, ensuring the supply chain's rapid response to market changes. Combining Bayesian dynamic linear models and convolutional neural networks enables dynamic adjustment of inventory levels, avoiding overstocking or shortages and improving inventory management efficiency. Based on multilayer perceptrons and fuzzy clustering algorithms, it achieves efficient supplier selection and ranking, ensuring stable supply chain operation and efficient supply. Through genetic algorithms and adaptive large-scale neighborhood search algorithms, it optimizes logistics routes, reduces transportation delays, and lowers transportation costs. Global data management and feedback mechanisms ensure synchronous operation among system units, improving overall supply chain collaboration efficiency. Based on logistics data and market trends, it enables timely adjustments to marketing strategies, enhancing customer experience and increasing order fulfillment rates.

[0068] Example 2: Please refer to Figure 2 This embodiment provides a specific implementation method for the demand forecasting and analysis unit in a data-driven intelligent logistics supply chain digital management system. The demand forecasting and analysis unit includes:

[0069] Short-term forecasting module: This module processes historical order data using a recurrent neural network (RNN). The RNN, with a specific number of time steps, progressively inputs time-series data, including historical order data, market fluctuation data, seasonal factors, and environmental variables. Through its recursive hidden state mechanism, the RNN combines the hidden state of the previous time step with the current input data to generate the prediction result for the next time step, thereby capturing short-term dependencies in the time-series data. These short-term dependencies reflect the correlation between historical order data in adjacent time steps. The RNN recursively models this correlation through multiple layers of hidden units, effectively identifying short-term fluctuation trends in order volume and generating accurate short-term forecasts accordingly.

[0070] The time step setting is based on the characteristics of order data and prediction requirements of specific application scenarios, defining the historical steps that RNN refers to when processing time series data. The time step is usually adjusted according to the fluctuation frequency and periodicity of order data, and can be determined in the following ways: (1) Data feature analysis: Through statistical analysis of historical order data, its fluctuation cycle and short-term change pattern are identified, and then the appropriate time step is determined. For example, if the order data shows daily fluctuations, the time step can be set to several days within a working cycle. (2) Experimental verification: By comparing the prediction effects of different time steps through experiments, the optimal time step that can improve training efficiency while ensuring prediction accuracy is selected. Usually, the time step is gradually increased until the prediction accuracy no longer improves significantly. (3) Dynamic adjustment: In some complex application scenarios, the time step can be dynamically adjusted based on system feedback. When order demand fluctuates frequently, the time step can be shortened to enhance the response to short-term changes; when order fluctuations tend to be stable, the time step can be appropriately increased to improve the prediction accuracy of long-term trends.

[0071] The short-term order forecasts generated by RNNs include order demand volume, demand fluctuation trends, and related market changes over a future period. These forecasts are calculated through the output layer of the recurrent neural network, the hidden layer states are processed based on activation functions, and the model parameters are optimized using the Adam optimizer, thereby improving prediction accuracy and model training efficiency.

[0072] 2. Medium- to Long-Term Forecasting Module: This module uses a Long Short-Term Memory (LSTM) network to analyze historical order data and market fluctuation data, capturing long-term dependencies in time series. The LSTM uses its gating mechanism: an input gate to select current input information, a forget gate to adjust the weight of historical data on the current forecast, and an output gate to generate the forecast result. This module handles demand forecasts over longer time spans to address long-term fluctuation trends in the supply chain.

[0073] The parameters of LSTM include the number of hidden layer units, typically set to 128 or 256, with an appropriate value chosen based on the data size. The optimization strategy uses the Adam optimizer, adjusting the learning rate to ensure the model can adapt to changes in different data distributions, thereby improving prediction accuracy. Input data includes short-term predictions generated by a recurrent neural network (RNN), historical order data, market trend data, and environmental factors, which are combined to generate medium- to long-term demand forecasts.

[0074] The medium- to long-term forecast results are used in the inventory optimization unit, logistics route optimization unit, and supplier management unit. Specifically, the forecast results generated by LSTM are applied to the convolutional neural network (CNN) in the inventory optimization unit to help identify long-term trends in inventory data and optimize inventory allocation. Furthermore, the forecast results support the adaptive large-scale neighborhood search algorithm (ALNS) in the logistics route optimization unit, providing a basis for long-term logistics resource scheduling and ensuring the rationality of logistics planning and efficient resource allocation. Simultaneously, the medium- to long-term forecast results guide the multilayer perceptron (MLP) and fuzzy clustering algorithms in the supplier management unit, ensuring the long-term stability and adaptability of supplier selection and order allocation.

[0075] 3. Attention Mechanism Module: This module weights the medium- to long-term forecast data output by the Long Short-Term Memory (LSTM) network. By assigning different weights to the input data at different time steps, it dynamically adjusts the model's focus on key nodes in the time series. This module is suitable for capturing seasonal fluctuations and sudden market events, thereby improving the accuracy of demand forecasting and the system's responsiveness.

[0076] Attention mechanisms employ two main algorithms for weighting: (1) Additive attention, which assigns weights to each time step by calculating the similarity between the hidden state at the current time step and historical data. The parameters of this algorithm include the dimension of the hidden units, typically set to 64 or 128, to adjust the model's sensitivity to more distant time steps; and (2) Multiplicative attention, which calculates weights through dot product operations and is suitable for large-scale data processing. The parameters of the multiplicative mechanism include the learning rate and the dot product dimension, where the learning rate is typically set to 0.001 or 0.0005, and the dot product dimension is set to 32 or 64 to control the precision of the attention weights and the training speed.

[0077] Different algorithms and parameter settings have a significant impact on model results. Additive attention mechanisms perform well in the face of subtle seasonal variations, while multiplicative attention mechanisms are more suitable for handling large-scale historical data. The learning rate setting directly affects the model's convergence speed and its ability to respond to new data. Higher learning rates (e.g., 0.001) are suitable for rapidly changing market environments, while lower learning rates (e.g., 0.0005) help improve model stability.

[0078] The attention-weighted prediction results are then fed into the inventory optimization unit and the logistics route optimization unit to further enhance the decision-making capabilities of supply chain management. Specifically, the convolutional neural network (CNN) in the inventory optimization unit uses the weighted data to identify long-term trends in inventory data, ensuring flexible adjustments to inventory allocation during demand fluctuations. The adaptive large-scale neighborhood search algorithm (ALNS) in the logistics route optimization unit prioritizes areas or time periods with significant demand fluctuations based on the weighted prediction results, thereby reducing waste of logistics resources and transportation delays.

[0079] Through dynamic weighted processing of the attention mechanism, the system can significantly improve its adaptability to market changes and ensure that each unit of the supply chain makes accurate responses at critical time points.

[0080] 4. Feedback Module: This module collects and processes real-time feedback data from the inventory optimization unit, logistics route optimization unit, and supplier management unit. This feedback data includes operational parameters such as inventory levels, logistics delays, and supplier response speeds. The feedback module transmits this data to various modules within the demand forecasting and analysis unit, dynamically adjusting the parameters of the recurrent neural network (RNN), long short-term memory network (LSTM), and attention mechanism to optimize forecast accuracy and system response efficiency.

[0081] The core functions of the feedback module include data collection, dynamic parameter adjustment, and feedback response. Its parameter settings mainly include the frequency and granularity of feedback data. The frequency of feedback data can be set hourly or daily, depending on the system's real-time requirements; the feedback granularity is selected based on the operational complexity of each unit in the supply chain, typically at the order level or inventory level.

[0082] The feedback module uses two algorithms for processing feedback data: (1) an adaptive filtering algorithm, which adjusts model parameters recursively and is suitable for fast-response scenarios. The key parameters of this algorithm are the filter order and the adjustment step size. The filter order is usually set to 3 or 5, and the adjustment step size is 0.01 to 0.05 to control the sensitivity of the model to short-term fluctuations; (2) a Bayesian optimization algorithm, which is used to dynamically adjust model parameters in complex scenarios, especially when there are large fluctuations in the system. The main parameters of this algorithm are the sampling frequency and the confidence interval of the objective function. The sampling frequency is set to sample once every 10 minutes, and the confidence interval is set to 95% to 99% to ensure that the prediction accuracy is maintained when there are large market changes.

[0083] Different algorithms and parameter settings have a significant impact on the results of the feedback module. For example, adaptive filtering algorithms can prioritize adjusting inventory holding strategies and improve system response speed when order demand fluctuates rapidly, while Bayesian optimization algorithms are suitable for complex, long-term fluctuation scenarios, ensuring optimized logistics routes and resource scheduling when dealing with supplier delivery delays or logistics bottlenecks.

[0084] The optimization results of the feedback module directly impact the inventory optimization unit and the logistics route optimization unit. Specifically, when the logistics route optimization unit reports frequent transportation delays, the feedback module instructs the attention mechanism to increase the weight of that time period, improving the model's ability to identify abnormal demand. Simultaneously, the feedback module adjusts the time steps of the RNN and the forget gate weights of the LSTM to ensure more accurate capture of market fluctuation trends in future predictions, optimizing the overall responsiveness and resource utilization of the supply chain.

[0085] Example 3: Please refer to Figure 3 In this embodiment, the inventory optimization unit includes a data acquisition module, an inventory forecasting module, an inventory adjustment module, a feedback processing module, and an inventory data module. These modules work closely together, combining Bayesian Dynamic Linear Model (BDLM) and Convolutional Neural Network (CNN) algorithms to achieve real-time prediction and dynamic adjustment of inventory status. The algorithms of the inventory optimization unit provide positive feedback to the algorithms of the supplier management unit and the logistics route optimization unit, ensuring the efficient operation of the entire supply chain system.

[0086] 1. Data Acquisition Module: This module acquires data in real-time from various units within the supply chain. Input data includes historical order data, real-time order volume, market fluctuation data, supplier delivery information, and logistics status data. This data is standardized before being transmitted to the inventory forecasting and inventory adjustment modules. Standardization ensures data consistency and the accuracy of subsequent calculations. Output data consists of standardized order data, supplier delivery information, and logistics status information, which are then input to the inventory forecasting and inventory adjustment modules. The data acquisition module optimizes data processing efficiency by adjusting the data update frequency (e.g., hourly or daily) and the granularity of data acquisition (order level, inventory level, logistics level). Clustering algorithms are introduced to filter and classify data, reducing redundant data and improving the overall system response speed.

[0087] 2. Inventory Forecasting Module: This module combines a Bayesian Dynamic Linear Model (BDLM) and a Convolutional Neural Network (CNN) for inventory forecasting. Input data comes from standardized historical order data, real-time order data, market fluctuation information, and logistics status data obtained from the data acquisition module. BDLM is used for short-term inventory demand forecasting, while CNN is used for long-term inventory trend analysis. Output data includes short-term inventory forecast results (based on BDLM) and medium-to-long-term inventory forecast results (based on CNN), which are passed to the inventory adjustment module for dynamic adjustment of inventory levels and to the supplier management unit and logistics route optimization unit to optimize supplier selection and logistics route planning. BDLM uses a recursive filtering algorithm to adjust model parameters in real time, adapting to rapid changes in short-term inventory demand and improving the system's sensitivity to short-term fluctuations. CNN uses a multi-layer convolutional structure to extract long-term trends from historical data and identify long-term inventory change patterns. The number of convolutional layers can be set to 4 to 6, and the pooling layer size can be set to 2x2 or 3x3. The optimized model can more accurately capture long-term features in inventory data, thereby improving the accuracy of long-term forecasts.

[0088] 3. Inventory Adjustment Module: Based on the short-term and medium-to-long-term forecasts generated by the inventory forecasting module, this module dynamically adjusts inventory levels to adapt to current demand. Input data comes from the short-term inventory forecast (based on BDLM) and medium-to-long-term inventory forecast (based on CNN) of the inventory forecasting module, including specific inventory demand forecasts and inventory fluctuation trends. Output data comes from the inventory adjustment plan, including inventory replenishment and reduction schemes; and is transmitted to the supplier delivery plan module in the supplier management unit and the resource scheduling module in the logistics route optimization unit. The inventory adjustment module dynamically adjusts inventory based on parameters such as inventory safety level, adjustment frequency, and adjustment threshold. The inventory safety level can be set to 10%-20%, and the adjustment frequency can be set to daily or hourly. By setting dynamic adjustment strategies, the inventory adjustment module can quickly respond to inventory shortages or surpluses in the short term, optimizing inventory turnover.

[0089] 4. Feedback Processing Module: This module collects feedback data from the supplier management unit and the logistics route optimization unit, such as supplier delivery times, logistics execution status, and inventory turnover rates. This feedback data is used to adjust the model parameters of the inventory forecasting and inventory adjustment modules in real time to ensure efficient supply chain operation. Input data comes from feedback information from the supplier management unit and the logistics route optimization unit, including supplier delivery data, logistics route status, and inventory execution data. Output data includes optimized inventory forecasting and adjustment strategies, which are then transmitted to the inventory forecasting and inventory adjustment modules.

[0090] This module employs adaptive filtering and Bayesian optimization algorithms to process feedback data. The adaptive filtering algorithm dynamically adjusts inventory strategies; the filter order can be set to 3 or 5, and the adjustment step size can be set to 0.01 to 0.05. The Bayesian optimization algorithm ensures accurate forecasting even when demand fluctuates significantly by adjusting the feedback frequency and confidence interval (95%-99%). Through dynamic optimization of the feedback processing module, the system can adaptively adjust the parameters of each module, thereby improving the overall efficiency of supply chain management.

[0091] Example 4: Please refer to Figure 4 This embodiment provides one possible implementation of a supplier management unit, which includes a supplier data acquisition module, a supplier analysis module, a supplier classification and ranking module, and a supplier feedback processing module. Through collaborative work with the inventory optimization unit and the logistics route optimization unit, it dynamically optimizes supplier selection and order allocation strategies, improving the efficiency and responsiveness of supply chain management. The close collaboration between the supplier management unit and the inventory optimization and logistics route optimization units, through a proactive feedback mechanism, enhances the overall efficiency and flexibility of the supply chain.

[0092] 1. Supplier Data Acquisition Module: This module collects supplier data in real time from various units of the supply chain, including delivery time, order completion rate, inventory demand data, and logistics status information. Through data standardization, it ensures the consistency and accuracy of input data, providing foundational data support for subsequent modules. Input data includes supplier delivery time, order completion rate, inventory status, and logistics status data. Output data includes standardized supplier data and is transmitted to the supplier analysis module and the supplier classification and ranking module. The data update frequency can be set to hourly or daily according to actual needs, suitable for different supply chain complexities. Data filtering and standardization ensure data quality and processing efficiency.

[0093] 2. Supplier Analysis Module: Employing a Multilayer Perceptron (MLP) algorithm, this module generates supplier priorities by analyzing historical delivery records, response speed, and delivery stability. Input data includes supplier delivery time, historical performance, order completion rate, and inventory status. Output data includes the supplier priority ranking results, which are then passed to the supplier classification and ranking module. The number of hidden layer units in the MLP can be set to 128 or 256, and the learning rate can be set to 0.001. More hidden units improve the model's ability to identify complex supplier features, while a reasonable adjustment of the learning rate can achieve a balance between accuracy and efficiency.

[0094] 3. Supplier Classification and Ranking Module: This module uses fuzzy clustering to classify suppliers and ranks them based on their response speed, delivery stability, and inventory requirements. Input data comes from the supplier analysis module, including supplier priority ranking results, response speed, and delivery stability. Output data includes the supplier classification results and is passed to the inventory optimization unit and logistics route optimization unit. The fuzzy clustering algorithm subdivides suppliers based on the number of clusters (3-5 clusters) and a classification threshold (0.05), ensuring that each supplier is accurately classified. This classification optimizes order allocation strategies and improves supply chain response speed and resource allocation efficiency.

[0095] 4. Supplier Feedback Processing Module: This module collects feedback information from the inventory optimization and logistics route optimization units to dynamically adjust supplier selection and ranking strategies. Input data includes supplier delivery time, inventory turnover rate, and logistics execution status. Output data includes the optimized supplier selection and ranking results, which are fed back to the supplier analysis module. An adaptive filtering algorithm is used to adjust supplier ranking and delivery strategies in real time. The filter order (can be set to 3 or 5) and adjustment step size (can be set to 0.01 to 0.05) ensure the timeliness and accuracy of feedback adjustments. The Bayesian optimization algorithm dynamically adjusts the sampling frequency and confidence interval (95% to 99%) of the feedback data to ensure the efficiency and adaptability of supplier management in response to market fluctuations.

[0096] Through close collaboration among its modules, this embodiment enables the supplier management unit to rapidly adjust its supplier selection strategy in response to changes in order demand, inventory levels, and logistics status, ensuring efficient supply chain operation. The MLP algorithm and fuzzy clustering algorithm effectively improve the response speed of each link in the supply chain by optimizing supplier classification and ranking, while the adaptive filtering algorithm and Bayesian optimization algorithm ensure that the system can adaptively adjust its supplier strategy to achieve optimal resource allocation in a complex and ever-changing supply chain environment.

[0097] Example 5: Please refer to Figure 5 In this embodiment, the logistics route optimization unit consists of a logistics data acquisition module, a route planning module, a resource scheduling module, and a route feedback module. These modules work collaboratively, combining genetic algorithms, adaptive large-scale neighborhood search (ALNS) algorithms, and Dijkstra's algorithm to achieve dynamic optimization and efficient scheduling of logistics routes. The algorithms of each module, along with those of the inventory optimization unit and the supplier management unit, create positive feedback loops, ensuring improved overall supply chain efficiency and resource utilization.

[0098] 1. Logistics Data Acquisition Module: Responsible for acquiring real-time logistics-related data from the supply chain system. This data includes inventory status, demand forecast information, traffic conditions, and vehicle status, etc. After standardization, it is transmitted to the route planning and resource scheduling modules. Input data includes inventory status, demand forecast information, traffic conditions, and vehicle status data. Output data is standardized logistics status data, which is transmitted to the route planning and resource scheduling modules.

[0099] The data collection frequency can be set to hourly or minutely as needed to ensure the real-time nature and accuracy of logistics status data. This module improves data quality and processing efficiency through data filtering and standardization, ensuring that subsequent modules' route planning and resource scheduling can make decisions based on accurate data.

[0100] 2. Path Planning Module: First, a genetic algorithm is used to generate an initial set of path solutions, representing multiple possible paths from the starting point to the destination. The genetic algorithm, simulating biological evolution, continuously optimizes the path solution set through crossover and mutation operations. The generated initial path solution set is then fed as input to Dijkstra's algorithm. Dijkstra's algorithm further calculates the local optima for each path based on these initial solutions, outputting the final shortest path.

[0101] The role of the genetic algorithm in the path planning module is to provide an initial set of path solutions for Dijkstra's algorithm. It generates random initial solutions and optimizes the solution set through an evolutionary process. Population Initialization: The genetic algorithm first generates a set of random path solutions (called the population), where each solution corresponds to a feasible path from the starting point to the destination. These paths are based on input data from the logistics data acquisition module, taking into account current inventory status, traffic conditions, and vehicle availability. Population Size: The size of the path solution set is typically set to 50 to 100; a larger population helps cover more path selection space. Fitness Evaluation: The fitness of each path solution is evaluated based on its path length, traffic conditions, and estimated transport time. A higher fitness path solution indicates that it is closer to the optimal solution. The fitness function is mainly calculated using the following parameters: Path length: the total distance of the path; Transport time: the estimated transport time based on current traffic conditions; Resource consumption: including vehicle fuel consumption, personnel costs, etc. Selection Operation: The genetic algorithm selects path solutions with higher fitness from the current population to enter the next generation through roulette wheel selection or tournament selection, thereby improving the average quality of the path solution set. Crossover: The crossover operation swaps parts of two parent path solutions to generate new path solutions. The crossover rate is typically set to 0.7 to 0.9. Crossover ensures information exchange between different paths, promoting path optimization. Mutation: To avoid getting trapped in local optima, the genetic algorithm mutates some path solutions, changing some nodes in the path. The mutation rate is typically set to 0.01 to 0.05. A lower mutation rate ensures stable search, while a higher mutation rate helps the search space escape local optima. Iterative Evolution: The genetic algorithm continuously optimizes the path solution set through multiple generations of iteration. This is typically set to 10 to 20 generations until a high-quality set of path solutions is found.

[0102] Dijkstra's algorithm, based on the initial path solution set generated by the genetic algorithm, further calculates the locally optimal path for each path solution, ensuring the selection of the shortest path under specified traffic and transportation conditions. Path Update: Dijkstra's algorithm expands the path step-by-step by sequentially selecting nodes from the starting point, assigning the shortest path distance to each node. During each calculation, Dijkstra's algorithm updates the shortest distance from the starting point to the current node and records the optimal path. Shortest Path Calculation: Based on the initial path of each solution, Dijkstra's algorithm calculates the shortest transportation time and path length for each path under actual traffic conditions. For example, in urban logistics, Dijkstra's algorithm prioritizes smooth-flowing roads, bypassing congested areas to reduce transportation time. Final Path Output: The path solution optimized by Dijkstra's algorithm is ultimately output as the shortest path scheme, including the sequence of nodes traversed, the total distance, and the estimated transportation time. These results are passed to the resource scheduling module for scheduling logistics resources.

[0103] The genetic algorithm in the path planning module generates multiple feasible path solutions through a large-scale global search, avoiding the local optimum trap caused by the initial selection. Dijkstra's algorithm, on the other hand, performs refined local optimum calculations for each path solution, ensuring that each path is the shortest path under the current traffic conditions. The crossover and mutation operations of the genetic algorithm can quickly handle the complexity of logistics networks (such as multi-node, multi-path selection problems). Dijkstra's algorithm can dynamically update path selections, responding to changes in traffic conditions in real time, ensuring the timeliness and efficiency of transportation.

[0104] The resource scheduling module is responsible for dynamically scheduling vehicles and other logistics resources based on the route plans generated by the route planning module and the current status of logistics resources. The resource scheduling module employs an Adaptive Large-Scale Neighborhood Search (ALNS) algorithm to ensure efficient resource scheduling under complex transportation demands and resource constraints, reducing logistics bottlenecks and optimizing resource utilization.

[0105] 1. Input data includes the optimal route plan from the route planning module, including route planning data such as origin, destination, nodes passed through, total distance and estimated transportation time; and transportation resources such as the real-time location, load, availability and current transportation task status of the transport vehicles.

[0106] 2. Output data: This includes the optimized resource scheduling plan, which includes the allocation of specific transportation tasks, vehicle route selection, and scheduling time arrangements, and is transmitted to the route feedback module and each execution unit.

[0107] 3. The specific process is as follows:

[0108] (1) Initial solution generation: Based on the route plan output by the route planning module, an initial scheduling plan is generated. The initial plan is initially allocated according to the current load, availability and transportation demand of the transport vehicles.

[0109] (2) Neighborhood Search: ALNS explores different scheduling schemes by searching multiple neighborhoods to find the optimal resource allocation method. The definition of a neighborhood is based on several conditions, including: considering the vehicle's load capacity, optimizing the vehicle load neighborhood for cargo allocation by dynamically adjusting the load of each vehicle; adjusting the time window neighborhood for vehicle arrival times at each node based on supplier delivery times and order priorities; and dynamically optimizing the vehicle's transportation route by combining the route planning results to ensure the optimal route neighborhood for both transportation cost and time.

[0110] (3) Local adjustment and global optimization: During the search process, ALNS gradually improves the scheduling plan by adjusting local neighborhoods, such as exchanging tasks between vehicles and adjusting the scheduling order. Through global search, it ensures that the global optimal solution is found under the condition of limited resources.

[0111] (4) Parameter settings:

[0112] Neighborhood size: The size of the neighborhood determines the breadth of the search space. A larger neighborhood can cover more resource allocation schemes and is suitable for complex transportation tasks; a smaller neighborhood helps improve computational efficiency and is suitable for simpler scheduling tasks.

[0113] Search depth: The search depth determines the number of iterations the algorithm performs in each neighborhood. A higher search depth (e.g., 50-100 iterations) is suitable for large-scale logistics networks, ensuring that the optimal solution is found in complex scenarios.

[0114] Adjustment rules: including vehicle load limits, transportation time windows, etc., to ensure that all constraints are met during the scheduling process.

[0115] This module utilizes the Adaptive Large Scale Neighborhood Search (ALNS) algorithm for dynamic scheduling, enabling the resource scheduling module to optimize the utilization of transportation resources in real time, reducing empty load rates and resource waste. By combining neighborhood search with local adjustments, the system can effectively respond to dynamic changes in transportation demand, ensuring efficient logistics scheduling in complex supply chain environments.

[0116] 4. Route Feedback Module: This module's input data includes transportation time (actual transportation time at each node, including the actual travel time of vehicles from the origin to each intermediate node and the final destination); resource utilization rate (vehicle load utilization rate, actual transported cargo volume, and vehicle empty load rate during transportation); transportation delays (monitoring any transportation delays relative to the planned time); and other logistics statuses (real-time data such as road traffic conditions, vehicle location, and inventory status). Output data includes feedback optimization data (optimized route planning adjustment suggestions, including suggestions for reallocating transport vehicles, changing transportation time, or adjusting transportation routes, which are transmitted to the route planning module and resource scheduling module); and improvement plans (specific improvement plans proposed to address problems in logistics execution, which are fed back to the inventory optimization unit and supplier management unit to adjust inventory strategies and supplier selection schemes).

[0117] The specific process is as follows:

[0118] (1) Feedback Data Collection: The route feedback module collects data from each stage of the transportation process, including transportation time, resource utilization, and transportation delays. This data mainly comes from onboard monitoring equipment, real-time monitoring information from the logistics management system, and status reports provided by suppliers and logistics companies.

[0119] (2) Feedback Analysis and Processing: The module uses adaptive filtering and Bayesian optimization algorithms to analyze and process the collected feedback data. The adaptive filtering algorithm is mainly used for real-time monitoring and rapid adjustment of transportation time and resource utilization; the Bayesian optimization algorithm is used to adjust the overall logistics route planning and resource allocation scheme according to the transportation delay situation to ensure the optimal logistics execution effect.

[0120] Adaptive filtering algorithms correct problems caused by logistics delays or resource shortages by adjusting dynamic parameters in transportation task allocation (such as vehicle scheduling and transportation sequence). The filter order (e.g., 3rd to 5th order) and the adjustment step size (e.g., 0.01 to 0.05) affect the speed and accuracy of the feedback response.

[0121] Bayesian optimization algorithms dynamically adjust the planning strategy in the route planning module by analyzing problems such as transportation delays. Especially when traffic congestion or vehicle unavailability occurs, they prioritize the most urgent routes to ensure overall logistics efficiency. The setting of sampling frequency and confidence intervals (95% to 99%) ensures accurate feedback and optimization under different conditions.

[0122] Parameter settings:

[0123] (1) Feedback frequency: Based on the real-time nature of logistics needs, the feedback frequency can be set to hourly or real-time feedback to ensure that any delays or waste of resources during transportation can be captured and dealt with in a timely manner.

[0124] (2) Feedback granularity: The granularity of data collection can be refined to the order level or vehicle level. The level of detail of the data can be adjusted according to the needs in order to more accurately optimize logistics routes and resource allocation.

[0125] In summary, this embodiment combines genetic algorithms, adaptive large-scale neighborhood search (ALNS), and Dijkstra's algorithm. The path planning module performs global search and combines it with local optimization to ensure optimal selection of logistics routes, reducing transportation time and costs. The resource scheduling module, through ALNS, efficiently schedules logistics resources under complex transportation demands and resource constraints, maximizing resource utilization and reducing empty loads and resource waste. The logistics data acquisition module works closely with the path feedback module to monitor and adjust logistics execution plans in real time. Based on feedback information such as transportation time, resource utilization, and traffic conditions, it dynamically optimizes path planning and resource scheduling, ensuring the efficiency and timeliness of the logistics process. The logistics path optimization unit works closely with the inventory optimization unit and supplier management unit to form a positive feedback mechanism, ensuring efficient coordination among logistics, inventory, and suppliers, and improving the overall operational efficiency of the supply chain.

[0126] Example 6: Please refer to Figure 6 In one possible implementation, the system integration and feedback unit includes a global data management module, a feedback optimization module, an intelligent collaboration module, and a dynamic monitoring module. The system integration and feedback unit coordinates the various algorithm modules to form an effective feedback mechanism, linking with units such as demand forecasting and inventory optimization to achieve global optimization. This ensures the overall optimization and dynamic adjustment of the supply chain system, including:

[0127] 1. Global Data Management Module: Responsible for acquiring and processing relevant data from various units (demand forecasting and analysis unit, inventory optimization unit, supplier management unit, and logistics route optimization unit). Input data includes demand forecast results, inventory levels, logistics execution data, and supplier response data. Output data includes standardized global data, which is then transmitted to the feedback optimization module and the intelligent collaboration module.

[0128] This module cleanses, formats, and standardizes the input data to ensure consistent format and high accuracy across all functional units. It uses clustering algorithms to perform preliminary data analysis, eliminating redundant information and improving data processing efficiency. The data collection frequency is set to real-time, hourly, or daily based on the actual needs of the supply chain, ensuring timely and effective data processing.

[0129] 2. Feedback Optimization Module: Based on the data processed by the global data management module, and considering inventory levels, logistics status, and actual supplier responses, this module adjusts the operating parameters of the demand forecasting and analysis unit, the inventory optimization unit, and the logistics route optimization unit. By analyzing the operating data of each unit, it dynamically adjusts the operating parameters to ensure the overall optimization of the supply chain system. Input data includes real-time feedback data such as inventory fluctuations, logistics delays, and supplier delivery status. Output data includes the optimized operating parameters and is transmitted to the inventory optimization unit, the logistics route optimization unit, and the supplier management unit.

[0130] This module incorporates Bayesian optimization algorithms to dynamically adjust the model parameters of each unit based on feedback data. For example, when the logistics route optimization unit reports a transportation delay, the feedback optimization module dynamically adjusts the route planning and resource scheduling strategies according to the degree of delay. By adjusting the parameters of the Bayesian Dynamic Linear Model (BDLM) in the inventory optimization unit and the Adaptive Large-Scale Neighborhood Search (ALNS) algorithm in the logistics route optimization unit, the system ensures that it can quickly respond to market changes and optimize the overall efficiency of the supply chain.

[0131] 3. Intelligent Collaboration Module: This module coordinates the operation of the demand forecasting and analysis unit, the inventory optimization unit, the supplier management unit, and the logistics route optimization unit by analyzing data provided by the global data management module. Input data includes demand forecasting results, logistics execution status, and supplier response data. Output data includes the optimized global scheduling plan, which is then transmitted to the inventory optimization unit, the logistics route optimization unit, and the supplier management unit.

[0132] This module employs a multi-objective genetic algorithm for global optimization, considering multiple objectives such as inventory holding costs, logistics and transportation time, and supplier response speed. By dynamically adjusting parameters such as population size, crossover rate, and mutation rate, it ensures that each unit can optimize synchronously according to global demand. The intelligent collaboration module also works in conjunction with the convolutional neural network (CNN) of the inventory optimization unit to ensure a balance between long-term and short-term demand forecasts.

[0133] 4. Dynamic Monitoring Module: This module monitors the supply chain's operational status in real time, identifying anomalies such as logistics delays, inventory imbalances, and supplier response delays. It provides early warning signals and prompts timely adjustments. Input data includes logistics delays, inventory imbalances, and supplier response status. Output data includes early warning signals and optimization suggestions, which are then transmitted to the feedback optimization module and the intelligent collaboration module. This module employs an adaptive neural network algorithm for trend prediction, forecasting potential system bottlenecks based on historical and real-time data, and ensuring supply chain stability through an early warning mechanism. The early warning frequency can be set to minutes or hours as needed, ensuring the system can respond quickly to anomalies.

[0134] Example 7: Please refer to Figure 7 This real-time logistics marketing unit provided in this example includes a logistics data monitoring module, a market trend prediction module, a user behavior analysis module, and a marketing feedback module.

[0135] 1. Logistics data monitoring module: Responsible for collecting and monitoring key data in the logistics link of the supply chain system in real time. The input data includes inventory levels, logistics path status, order completion status, and customer feedback information. Through data cleaning and standardization processing, the accuracy and timeliness of the data are ensured.

[0136] The monitored logistics data includes: Inventory levels: Reflect the current inventory status and its matching degree with market demand; Logistics path status: Based on the feedback from the logistics path optimization unit, monitor the execution of each transportation path, including transportation time, delay situation, etc. Order completion status: Track the execution progress and delivery status of orders in real time. Customer feedback: Collect customers' evaluations of logistics services and their satisfaction with orders.

[0137] The output data is standardized logistics status data, which is transmitted to the market trend prediction module and the user behavior analysis module to ensure that marketing strategies can be adjusted according to the real-time situation of logistics execution.

[0138] 2. Market trend prediction module: Based on historical sales data, market volatility data, and customer behavior data, combined with the real-time logistics data provided by the logistics data monitoring module, dynamically predict future market demand trends.

[0139] The input data includes historical sales data, market volatility data, logistics status data, customer feedback data, etc.

[0140] The output data includes predicted market trends and demand changes, which are transmitted to the user behavior analysis module and the marketing feedback module for formulating dynamic marketing strategies. This module uses a long short-term memory network (LSTM) to analyze the short-term fluctuations and long-term trends of market demand, combined with changes in logistics status, to ensure that the market prediction results have high timeliness and accuracy. The time step setting of LSTM can be adjusted according to the cycle of market volatility, for example, set to 7 days, 30 days, or 90 days to reflect seasonal fluctuations or long-cycle market changes.

[0141] 3. User behavior analysis module: Generate personalized marketing plans for different customer groups by analyzing customers' order history, purchase behavior, and logistics feedback data. The input data includes market demand data output by the market trend prediction module, customer order history, and logistics feedback data. The output data includes personalized marketing suggestions, which are transmitted to the marketing feedback module and the inventory optimization unit.

[0142] The user behavior analysis module employs a multilayer perceptron (MLP) algorithm, combining factors such as customers' historical purchase preferences and logistics service quality feedback to generate customized marketing strategies for different customer groups. The number of hidden layer units and the learning rate of the MLP are optimized based on the complexity of the customer group and the amount of data, ensuring that the model can accurately capture customer behavior patterns and improve the targeting and effectiveness of marketing strategies.

[0143] 4. Marketing Feedback Module: This module collects feedback data on marketing campaign execution, adjusts marketing strategies in real time, and feeds it back to the market trend forecasting and user behavior analysis modules. Input data includes marketing campaign results, customer feedback on logistics services, and order completion status. Output data includes optimized marketing strategies and market forecast data, fed back to the market trend forecasting and user behavior analysis modules. The feedback frequency is set to hourly or daily based on the marketing campaign's performance, ensuring that marketing strategies can be adjusted according to the latest market feedback. Through this feedback mechanism, the marketing feedback module transmits real-time customer data to the market trend forecasting module, ensuring that market forecasts and user behavior analysis can adjust strategies based on actual conditions.

[0144] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0145] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A data analysis-based intelligent logistics supply chain digital management system, characterized in that, The system comprises: a demand prediction and analysis unit, which analyzes historical orders and market fluctuation data through a recurrent neural network and a long short-term memory network, predicts short-term and medium-and-long-term demand, and outputs demand prediction data; an inventory optimization unit, which combines a Bayesian dynamic linear model and a convolutional neural network algorithm, adjusts inventory levels according to the demand prediction data output by the demand prediction and analysis unit; a supplier management unit, which is based on a multi-layer perceptron and a fuzzy clustering algorithm, receives inventory adjustment data from the inventory optimization unit, and selects, classifies and sorts suppliers according to supplier data; the unit also combines the delivery capacity and historical performance of suppliers to develop a supplier delivery plan and adjust order allocation strategies; a logistics path optimization unit, which uses a genetic algorithm and an adaptive large-scale neighborhood search algorithm to plan logistics paths based on the supplier selection results of the supplier management unit and real-time logistics data; a system integration and feedback unit, which receives data from the demand prediction and analysis unit, the inventory optimization unit, the supplier management unit and the logistics path optimization unit through a global data management and feedback mechanism, integrates and synchronizes the operation of each unit; a real-time logistics marketing unit, which receives feedback data from the system integration and feedback unit, adjusts logistics marketing strategies based on logistics data and market trends; The demand prediction and analysis unit comprises: a short-term prediction module, which generates short-term demand prediction results through a recurrent neural network and transmits the prediction results to the inventory optimization unit to adjust inventory levels; a medium-and-long-term prediction module, which generates medium-and-long-term demand prediction using a long short-term memory network, and transmits the output results of the long short-term memory network combined with the short-term prediction results of the recurrent neural network algorithm to the inventory optimization unit and the logistics path optimization unit to develop long-term inventory and logistics scheduling schemes; an attention mechanism module, which assigns different weights to the prediction data output by the long short-term memory network and transmits the weighted prediction data to the supplier management unit and the logistics path optimization unit; a feedback module, which collects the operation data of the inventory optimization unit, the logistics path optimization unit and the supplier management unit, and transmits feedback information to the recurrent neural network and the long short-term memory network to adjust model parameters according to actual operation; The long short-term memory network in the medium-and-long-term prediction module of the demand prediction and analysis unit comprises an input gate, a forget gate and an output gate, the input gate is used to select current input information, the forget gate is used to adjust the weight of historical data on current prediction, and the output gate is used to generate prediction results; The inventory optimization unit comprises a data acquisition module, a stock prediction module, an inventory adjustment module, and a feedback processing module; the data acquisition module is configured to acquire data from the demand prediction analysis unit, the supplier management unit, the logistics path optimization unit, and the inventory optimization unit, wherein the data comprises historical order data, real-time order volume, market fluctuation data, supplier delivery information, and logistics state data; the data is transmitted to the inventory prediction module and the inventory data module after standardization processing, thereby providing a basis for inventory prediction and adjustment; the stock prediction module is configured to perform inventory prediction in combination with a Bayesian dynamic linear model and a convolutional neural network; the Bayesian dynamic linear model is used for short-term inventory prediction, and the convolutional neural network is used for analyzing long-term inventory trends; the prediction result is transmitted to the inventory adjustment module, thereby providing a basis for short-term and long-term inventory regulation and control; the inventory adjustment module is configured to perform dynamic adjustment of the inventory level based on the prediction result of the inventory prediction module; the feedback processing module is configured to collect operation feedback information from the supplier management unit and the logistics path optimization unit, and transmit the operation feedback information to the inventory prediction module and the inventory adjustment module. The convolutional neural network in the inventory optimization unit comprises a plurality of convolutional layers and pooling layers; the convolutional layers are used for processing local patterns in inventory data, and the pooling layers are used for reducing the dimension of data and extracting long-term trend features of the inventory data.

2. The data analysis based intelligent logistics supply chain digital management system according to claim 1, characterized in that, The supplier management unit comprises a supplier data acquisition module, a supplier analysis module, a supplier classification and sorting module, and a supplier feedback processing module; the supplier data acquisition module is configured to acquire supplier-related data from each unit of the supply chain, including supplier delivery time, order completion rate, inventory demand data, and logistics state information; the acquired data is transmitted to the supplier analysis module and the supplier classification and sorting module after standardization processing; the supplier analysis module is configured to analyze the historical delivery records, response speed, and delivery stability of the supplier based on a multilayer perceptron algorithm, and generate a supplier priority; according to the analysis result, the supplier delivery plan is adjusted, and the supplier priority sorting result is transmitted to the supplier classification and sorting module; the supplier classification and sorting module is configured to classify and sort the suppliers by using a fuzzy clustering algorithm; according to the supplier response speed, delivery stability, and inventory demand, the supplier classification result is determined and transmitted to the inventory optimization unit and the logistics path optimization unit; the supplier feedback processing module is configured to collect and process operation data from other units of the supply chain, including supplier delivery time, logistics execution, and inventory turnover rate; the feedback data is transmitted to the supplier analysis module. 3.The data analysis based intelligent logistics supply chain digital management system according to claim 1, characterized in that, The logistics path optimization unit comprises: a logistics data acquisition module: for acquiring logistics related data, including inventory status, demand prediction information, traffic conditions and transportation vehicle state information; the acquired data is transmitted to the path planning module and the resource scheduling module after processing; the path planning module: based on genetic algorithm and Dijkstra algorithm to generate the optimal path scheme; the genetic algorithm is used to generate the initial path solution set, and the Dijkstra algorithm calculates the shortest path scheme based on the initial solution, and transmits the scheme to the resource scheduling module; the resource scheduling module: according to the path scheme generated by the path planning module, and using adaptive large-scale neighborhood search algorithm to schedule logistics resources; the path feedback module: for collecting the actual execution data of logistics path, including transportation time, resource utilization rate and transportation delay; the feedback data is transmitted to the path planning module and the resource scheduling module to adjust the path planning and resource scheduling scheme.

4. The data analysis based intelligent logistics supply chain digital management system according to claim 3, characterized in that, The initial path solution set of the genetic algorithm is set to 50 to 100. 5.The data analysis based intelligent logistics supply chain digital management system according to claim 1, wherein, The system integration and feedback unit comprises: a global data management module: for acquiring data from the demand prediction and analysis unit, the inventory optimization unit, the supplier management unit and the logistics path optimization unit; the acquired data is standardized and transmitted to the feedback optimization module and the intelligent collaboration module; the feedback optimization module: according to the data processed by the global data management module, combined with the actual response of inventory level, logistics state and supplier, adjusts the operating parameters of the demand prediction and analysis unit, the inventory optimization unit and the logistics path optimization unit; the intelligent collaboration module: through analyzing the data provided by the global data management module, coordinates the operation of the demand prediction and analysis unit, the inventory optimization unit, the supplier management unit and the logistics path optimization unit; the dynamic monitoring module: through monitoring the logistics delay, inventory imbalance and supplier response, identifies potential problems and provides early warning signals.

6. The data analysis based intelligent logistics supply chain digital management system according to claim 1, wherein, The real-time logistics marketing unit comprises: a logistics data monitoring module: for monitoring key data in the logistics process, including inventory level, logistics path state, order completion and customer feedback information; the monitored data is transmitted to the market trend prediction module and the user behavior analysis module; the market trend prediction module: based on historical sales data, market fluctuation data and logistics data, predicts future market demand trend; the prediction result is transmitted to the user behavior analysis module and the marketing feedback module; the user behavior analysis module: through analyzing the customer's order history, purchase behavior and logistics feedback data, generates marketing scheme for different customer groups; the generated marketing scheme is transmitted to the marketing feedback module and the inventory optimization unit; the marketing feedback module: for collecting the execution data of marketing activities and the feedback information of customers on logistics service; the feedback data is transmitted to the market trend prediction module and the user behavior analysis module to adjust the market prediction and marketing strategy.

Images