Logistics scheduling management system and method based on artificial intelligence

Abstract

The invention discloses a logistics scheduling management system and method based on artificial intelligence, and belongs to the technical field of artificial intelligence. The method comprises the following steps: acquiring multi-source data from each online link of international trade, and preprocessing the acquired multi-source data; generating a merchant order data set, a type order data set, a merchant order relation chain and a merchant similar order relation chain; analyzing the merchant order data set by using a support vector machine algorithm, calculating the logistics priority of each order, and generating a logistics priority order data set; in combination with the logistics priority of each order, a logistics transportation path is optimized, and the orders with high priorities are delivered preferentially; based on the merchant similar order relation chain, in combination with the type order data set, clustering analysis is carried out on similar goods, and a similar goods set is formed; forming a merchant relationship set by using the merchant order relationship chain; and based on the similar goods set and the merchant relationship set, generating a logistics binding scheme, and scheduling logistics vehicles or other transportation tools.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based logistics scheduling management system and method. Background Art

[0002] With the rapid development of globalization and e-commerce, logistics scheduling and management in international trade are becoming increasingly complex, especially when it comes to processing multi-source data and dynamic logistics scheduling. International trade typically involves multiple stages, including order generation, logistics tracking, and managing merchant partnerships. These stages generate massive amounts of heterogeneous data. Effectively collecting, cleaning, analyzing, and optimizing this data to improve logistics efficiency is a major technical challenge.

[0003] When it comes to logistics prioritization and route optimization, existing systems mostly rely on static rules or manual judgment, failing to dynamically adjust transportation plans based on real-time logistics data, which can easily lead to transportation delays and wasted resources. Existing technologies for batch transportation often rely on manual or simple fixed algorithms, failing to intelligently bundle shipments based on cargo type, order relationships, and inter-merchant partnerships. This can easily lead to wasted transportation resources and increased logistics costs. Existing logistics systems rarely consider long-term partnerships or transaction frequency between merchants, resulting in an inability to optimize logistics collaboration between merchants and a missed opportunity to improve transportation efficiency. Summary of the Invention

[0004] The purpose of the present invention is to provide a logistics scheduling management system and method based on artificial intelligence to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions: A logistics scheduling management method based on artificial intelligence, the method comprising the following steps: S100. Collect multi-source data from various online links of international trade, including transaction orders, transaction logistics, and transaction merchant relationship chains; clean and format the collected multi-source data, remove redundant information and abnormal data, and standardize the multi-source data; S200: Based on the transaction orders, the data are divided into a merchant order dataset and a type order dataset according to the transaction merchant and the product type; based on the merchant order dataset and the transaction merchant relationship chain, a merchant order relationship chain is generated; based on the type order dataset and the merchant order relationship chain, a merchant similar order relationship chain is generated; S300: Utilize the support vector machine algorithm to analyze the order urgency, cargo transportation distance, and destination in the merchant order dataset, calculate the logistics priority of each order, and generate a logistics priority order dataset. Based on the logistics priority of each order, optimize the logistics transportation path using the reinforcement learning algorithm Q-learning. Dynamically adjust the transportation plan based on the cargo's destination, real-time traffic conditions, and logistics costs, prioritizing the delivery of high-priority orders. S400. Based on the merchant order relationship chain and in combination with the type order data set, similar goods are clustered and analyzed to identify similar goods that can be transported together and form a similar goods set; using the merchant order relationship chain, merchants with order relationships are identified, and the transaction frequency and geographical location relationship between merchants are analyzed through the Dijkstra algorithm to form a merchant relationship set; based on the similar goods set and the merchant relationship set, a logistics bundling plan is generated to bundle merchants in the same region or logistics node with similar goods for transportation; according to the generated bundling transportation plan, logistics vehicles or other transportation tools are dispatched to arrange batch transportation of goods.

[0006] According to step S100, transaction orders are automatically collected from e-commerce platforms, B2B transaction websites, and enterprise management systems, including order numbers, transaction times, product details, transaction quantities, transaction prices, and information about both the buyer and seller. Transaction logistics are obtained through logistics management systems and third-party logistics service providers, including shipping times, transportation status, delivery routes, and estimated arrival times. Contact information and transaction history of each merchant are collected to form a merchant relationship chain. De-duplication is performed on the collected multi-source data to ensure that each data record is unique; outliers are identified and eliminated through Z-score analysis; data from different sources are formatted in a unified manner to ensure consistency in field names and data types; numerical data is converted to numerical format, and categorical data is converted to the corresponding coding format; multi-source data is fused, and the cleaned and formatted data is integrated into the data warehouse.

[0007] According to step S200, merchant information is extracted from the buyer and seller information in the transaction order, and orders are grouped according to merchant ID to form a merchant order dataset; product type information is extracted from the product details in the transaction order, and orders are grouped according to product type to generate a type order dataset; based on the merchant order dataset and combined with the transaction merchant relationship chain, a merchant order relationship chain is established; the merchant order relationship chain records the transaction relationships between different merchants and shows the mutual influence of each merchant's orders; the structure of the merchant order relationship chain is defined, where nodes represent merchants, edges represent transaction relationships, and each edge is attached with information from the transaction order; Using the type order dataset, we identify orders of similar goods in each merchant's order dataset, and determine the connections between similar orders by analyzing the merchant order relationship chain; we build a merchant similar order relationship chain to record the relationship between orders of similar goods and the merchants to which they belong.

[0008] Extract the details of each item from the type order dataset and associate it with the merchant order dataset. Use the item classification labels to compare the items in each merchant order dataset with the items in the type order dataset, and identify similar items based on the preset classification criteria. Extract the transaction records between merchants from the merchant order dataset and establish a preliminary merchant order relationship chain, treating each merchant as a node and the transaction records as edges connecting the nodes. Based on the transaction quantity and transaction price, the strength of the relationship between merchants is evaluated and represented in the form of a weighted graph. The weight of the edge represents the closeness of the transaction between merchants. The identified similar goods are combined with the merchant order relationship chain to construct a merchant similar order relationship chain. Each edge represents an order for similar goods and the relationship between the merchants to which the order belongs. The node of each merchant similar order relationship chain represents the merchant, and the edge represents the order relationship of similar goods. Each edge is attached with information in the transaction order.

[0009] According to step S300, the urgency score of the order is calculated based on the difference between the order creation time and the expected delivery time, generating the order urgency level. The smaller the difference, the higher the urgency level. The calculated transportation distance data is used as the cargo transportation distance in kilometers. The destination is encoded, specifically by using one-hot encoding to convert the destination into a numerical feature. Based on actual business needs, a logistics priority label is defined for each order. The order dataset is divided into a training set and a test set. A support vector machine model is selected and trained using the training set. Order urgency, cargo transportation distance, and destination are used as feature inputs, and priority labels are used as target outputs. Model parameters are adjusted to optimize classification accuracy. The trained support vector machine model is used to predict the test set to generate a logistics priority score for each order. The model's prediction effect is evaluated using accuracy, recall, and F1-score. The prediction results are combined with the original order data to generate a logistics priority order dataset, which contains order ID, order urgency score, cargo transportation distance, destination code, and predicted logistics priority.

[0010] The urgency of the order, the distance of the cargo transportation, and the destination are analyzed through scatter plots to view the relationship and distribution of the features and obtain visualization results. Based on the visualization results, it is determined whether the data is linearly separable. When the features show a linear separable trend, the linear SVM is selected as the support vector machine model. The linear SVM uses a linear decision boundary to divide different categories. When the features show a nonlinear distribution, the RBF kernel is selected as the support vector machine model. The RBF kernel can map the data to a high-dimensional space and find linear boundaries in the high-dimensional space.

[0011] According to step S300, the logistics transportation path is optimized by using the reinforcement learning algorithm Q-learning in combination with the logistics priority of each order. The specific steps are as follows: S301. Define a state space, where each state represents a specific transportation environment, including the current transportation route, cargo destination, real-time traffic conditions, transportation costs, and logistics priority of the current order. Each state is represented by a vector. S302: Define an action space, where actions represent selectable transportation paths or modes; set a reward mechanism to encourage Q-learning to optimize the path; S303. Create a Q table and initialize the state-action value function Q(s, a) to 0, which represents the expected reward for taking a specific action a in a specific state s. Use the ε-greedy strategy to select actions. In the early stages of training, randomly select actions to explore the environment. As learning deepens, gradually reduce the randomness and select the action with the highest current Q value. After taking an action, observe the changes in the environment, obtain the new state and reward, and update the Q value. The update formula is: ; in, is the learning rate, r is the immediate reward, is the discount factor, For the new state, is a new action; after several iterations, by continuously updating the Q value, the model's strategy for transport path selection is enhanced and the optimal path is gradually found.

[0012] According to step S400, a K-means algorithm is selected as a clustering algorithm to cluster similar goods; the clustering algorithm is executed to divide the similar goods into a plurality of clusters, each cluster representing a group of similar goods that can be transported together, forming a similar goods set; during the clustering process, the number of clusters k is determined by the elbow rule; Merchant transaction records are extracted from the merchant order relationship chain. Merchants are regarded as nodes in the network, and transaction records are regarded as edges connecting nodes. The weights of the edges are weighted based on the transaction frequency to form a weighted undirected graph. The Dijkstra algorithm is used to analyze the transaction frequency and geographical location relationship between merchants, and the distance between merchants is set as the weight to optimize the transportation path. The relationships between the obtained merchants are integrated to form a merchant relationship set.

[0013] Extract information from similar goods sets and merchant relationship sets to identify merchants in the same region or logistics node and their corresponding similar goods sets; determine bundling rules, including geographic location, goods characteristics, and transportation costs; for geographic location, select merchants in the same city or logistics node for bundling; for goods characteristics, ensure that the bundled goods are of the same or compatible types; and for transportation costs, prioritize bundles with low transportation costs; Based on the bundled transport rules, a logistics bundling plan is generated, each plan including: a merchant list, a cargo list, and a transportation route; wherein the merchant list is the merchant ID and location of the merchants participating in the bundled transport, the cargo list is the same type of cargo and quantity for the bundled transport, and the transportation route is the optimal transportation route from the starting point to each destination; according to the bundling plan, the transportation tools are evaluated and batch transportation is carried out; and a transportation plan is formulated, including the departure time, route, and arrival time.

[0014] Based on the merchant order chain, merchants with order connections are identified. Merchants with relatively close geographic locations are selected and added to the merchant list. Merchant IDs are extracted from order data, and merchant location information can be obtained from merchant registration information or location data in the logistics system. The Dijkstra algorithm is used to analyze the transaction frequency and cooperation strength between merchants, prioritizing merchants with a high order volume or cooperative relationships for logistics bundling.

[0015] During cluster analysis, the system aggregates similar goods from different merchants to form a similar goods set. The type of goods is determined by the classification label in the order data, while the quantity is determined by the detailed information in each order. The goods list includes the specific types of goods and the corresponding shipping quantities. Similar goods typically have similar transportation requirements, such as temperature, humidity, volume, and packaging method. Therefore, the goods list ensures that all bundled goods meet the same transportation conditions to ensure safe and efficient transportation.

[0016] Based on the merchant's geographic location and destination, we use a routing algorithm like Dijkstra's shortest path algorithm or A* to generate an optimal transportation route. This route takes into account traffic conditions, transportation distances, and the cargo needs of each merchant to minimize transportation time and costs. Taking into account the logistics priority of each order, we prioritize the delivery of high-priority orders as quickly as possible. When planning routes, we first visit the destinations of high-priority orders, and then schedule the delivery locations of lower-priority orders.

[0017] Evaluate the capacity of the transport vehicle based on the quantity and type of goods in your cargo list. Common transport vehicles include trucks, vans, and air freight containers. When selecting a vehicle, ensure that the vehicle's volume is sufficient to accommodate all bundled goods and meets transportation requirements, such as temperature and humidity control. Select the appropriate transport vehicle based on the distance of the transport route and the geographic location of the destination. For example, a van may be suitable for short-distance transport, while air or sea freight may be required for long, cross-border shipments.

[0018] An artificial intelligence-based logistics scheduling and management system, comprising: Multi-source data acquisition and processing module: includes a data acquisition unit and a data pre-processing unit. The data acquisition unit collects multi-source data from various online links of international trade, including transaction orders, transaction logistics, and transaction merchant relationship chains. The data pre-processing unit cleans and formats the collected multi-source data, removes redundant information and abnormal data, and standardizes the multi-source data. Data classification and relationship chain generation module: includes: a data classification unit, a merchant order relationship chain generation unit, and a merchant similar order relationship chain generation unit; wherein the data classification unit divides the transaction order into a merchant order data set and a type order data set according to the transaction merchant and the type of goods respectively; the merchant order relationship chain generation unit generates a merchant order relationship chain based on the merchant order data set and the transaction merchant relationship chain; and the merchant similar order relationship chain generation unit generates a merchant similar order relationship chain based on the type order data set and the merchant order relationship chain; Logistics Priority Calculation and Path Optimization Module: This module includes a logistics priority calculation unit and a logistics path optimization unit. The logistics priority calculation unit uses a support vector machine algorithm to analyze the order urgency, cargo transportation distance, and destination in the merchant order dataset, calculates the logistics priority of each order, and generates a logistics priority order dataset. The logistics path optimization unit combines the logistics priority of each order and optimizes the logistics transportation path through the reinforcement learning algorithm Q-learning. It dynamically adjusts the transportation plan based on the cargo's destination, real-time traffic conditions, and logistics costs, giving priority to the delivery of high-priority orders. Logistics bundling plan generation module: includes: similar goods clustering analysis unit, merchant relationship analysis unit, logistics bundling plan generation unit and transportation tool scheduling unit; among them, the similar goods clustering analysis unit is based on the merchant similar order relationship chain, combined with the type order data set, to cluster the similar goods, identify the similar goods that can be transported together, and form a similar goods set. The merchant relationship analysis unit uses the merchant order relationship chain to identify merchants with order relationships, and analyzes the transaction frequency and geographical location relationship between merchants through the Dijkstra algorithm to form a merchant relationship set. The logistics bundling plan generation unit generates a logistics bundling plan based on the similar goods set and the merchant relationship set, and bundles merchants in the same area or logistics node with similar goods for transportation. The transportation tool scheduling unit dispatches logistics vehicles or other transportation tools according to the generated bundled transportation plan to arrange batch transportation of goods.

[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention introduces a support vector machine algorithm to intelligently evaluate the logistics priority of orders based on multi-dimensional features such as order urgency, transportation distance, and destination. Compared with traditional static rules, this invention can dynamically adjust priorities based on real-time data to ensure the timely delivery of critical orders.

[0020] 2. The present invention uses a reinforcement learning algorithm to optimize routes based on logistics priorities and real-time traffic conditions, dynamically adjust transportation plans, improve transportation efficiency, and reduce transportation costs. Compared with static planning, the present invention can adapt to complex transportation environments and achieve more flexible scheduling.

[0021] 3. By analyzing similar goods through clustering algorithms and combining them with the merchant order relationship chain, the present invention can intelligently identify similar goods that can be transported together and merchants with cooperative relationships, generate bundled transportation plans, maximize the use of logistics resources, and improve batch transportation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a schematic diagram of the steps of a logistics scheduling management method based on artificial intelligence of the present invention; Figure 2 This is a system structure diagram of an artificial intelligence-based logistics scheduling management system of the present invention. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution. According to one embodiment of the present invention, Figure 1 A schematic diagram of a logistics scheduling management method based on artificial intelligence is shown, and a logistics scheduling management method based on artificial intelligence includes the following steps: S100. Collect multi-source data from various online links of international trade, including transaction orders, transaction logistics, and transaction merchant relationship chains; clean and format the collected multi-source data, remove redundant information and abnormal data, and standardize the multi-source data; S200: Based on the transaction orders, the data are divided into a merchant order dataset and a type order dataset according to the transaction merchant and the product type; based on the merchant order dataset and the transaction merchant relationship chain, a merchant order relationship chain is generated; based on the type order dataset and the merchant order relationship chain, a merchant similar order relationship chain is generated; S300: Utilize the support vector machine algorithm to analyze the order urgency, cargo transportation distance, and destination in the merchant order dataset, calculate the logistics priority of each order, and generate a logistics priority order dataset. Based on the logistics priority of each order, optimize the logistics transportation path using the reinforcement learning algorithm Q-learning. Dynamically adjust the transportation plan based on the cargo's destination, real-time traffic conditions, and logistics costs, prioritizing the delivery of high-priority orders. S400. Based on the merchant order relationship chain and in combination with the type order data set, similar goods are clustered and analyzed to identify similar goods that can be transported together and form a similar goods set; using the merchant order relationship chain, merchants with order relationships are identified, and the transaction frequency and geographical location relationship between merchants are analyzed through the Dijkstra algorithm to form a merchant relationship set; based on the similar goods set and the merchant relationship set, a logistics bundling plan is generated to bundle merchants in the same region or logistics node with similar goods for transportation; according to the generated bundling transportation plan, logistics vehicles or other transportation tools are dispatched to arrange batch transportation of goods.

[0025] According to step S100, transaction orders are automatically collected from e-commerce platforms, B2B transaction websites, and enterprise management systems, including order numbers, transaction times, product details, transaction quantities, transaction prices, and information about both the buyer and seller. Transaction logistics are obtained through logistics management systems and third-party logistics service providers, including shipping times, transportation status, delivery routes, and estimated arrival times. Contact information and transaction history of each merchant are collected to form a merchant relationship chain. De-duplication is performed on the collected multi-source data to ensure that each data record is unique; outliers are identified and eliminated through Z-score analysis; data from different sources are formatted in a unified manner to ensure consistency in field names and data types; numerical data is converted to numerical format, and categorical data is converted to the corresponding coding format; multi-source data is fused, and the cleaned and formatted data is integrated into the data warehouse.

[0026] According to step S200, merchant information is extracted from the buyer and seller information in the transaction order, and orders are grouped according to merchant ID to form a merchant order dataset; product type information is extracted from the product details in the transaction order, and orders are grouped according to product type to generate a type order dataset; based on the merchant order dataset and combined with the transaction merchant relationship chain, a merchant order relationship chain is established; the merchant order relationship chain records the transaction relationships between different merchants and shows the mutual influence of each merchant's orders; the structure of the merchant order relationship chain is defined, where nodes represent merchants, edges represent transaction relationships, and each edge is attached with information from the transaction order; Using the type order dataset, we identify orders of similar goods in each merchant's order dataset, and determine the connections between similar orders by analyzing the merchant order relationship chain; we build a merchant similar order relationship chain to record the relationship between orders of similar goods and the merchants to which they belong.

[0027] Extract the details of each item from the type order dataset and associate it with the merchant order dataset. Use the item classification labels to compare the items in each merchant order dataset with the items in the type order dataset, and identify similar items based on the preset classification criteria. Extract the transaction records between merchants from the merchant order dataset and establish a preliminary merchant order relationship chain, treating each merchant as a node and the transaction records as edges connecting the nodes. Based on the transaction quantity and transaction price, the strength of the relationship between merchants is evaluated and represented in the form of a weighted graph. The weight of the edge represents the closeness of the transaction between merchants. The identified similar goods are combined with the merchant order relationship chain to construct a merchant similar order relationship chain. Each edge represents an order for similar goods and the relationship between the merchants to which the order belongs. The node of each merchant similar order relationship chain represents the merchant, and the edge represents the order relationship of similar goods. Each edge is attached with information in the transaction order.

[0028] In this example, a transaction order dataset is used, which contains the following information: Merchant Information: Each buyer and seller in an order has a unique merchant ID, which represents the merchant involved in the transaction.

[0029] Goods details: Each order records in detail information such as goods type, quantity, unit price, etc.

[0030] Transaction records: Orders include the quantity and amount of transactions, representing the transaction relationship between merchants.

[0031] Using a merchant order dataset, we aggregate transaction records between merchants, treating each merchant as a node and transaction relationships as edges. We determine the strength of relationships between merchants based on the number and amount of transactions. Based on the number and amount of transactions, the resulting merchant order relationship chain is as follows: A and B have two transactions with a total transaction amount of 11,240 yuan, resulting in a transaction relationship strength of 2; A and C have one transaction with a total transaction amount of 4,000 yuan; and B and C have one transaction with a total transaction amount of 4,400 yuan.

[0032] The merchant order relationship chain can be viewed as the following weighted graph: A↔B (weight = 2 transactions, amount = 11,240 yuan), A↔C (weight = 1 transaction, amount = 4,000 yuan), B↔C (weight = 1 transaction, amount = 4,400 yuan).

[0033] Using product classification labels, we compare goods in merchant orders, identify similar goods, and generate a similar goods set. For example, goods in the electronic product category are grouped together, forming a similar goods set for electronic products. Combining the similar goods set with the merchant order relationship chain, we construct a similar merchant order relationship chain. Each edge represents the relationship between orders for similar goods and their respective merchants. For example, if there are two orders for electronic products between merchants A and B, and one order for electronic products between merchants B and C, the similar merchant order relationship chain is as follows: A↔B (Goods: Electronic Products, Number of Transactions = 2, Quantity = 22), B↔C (Goods: Electronic Products, Number of Transactions = 1, Quantity = 8).

[0034] According to step S300, the urgency score of the order is calculated based on the difference between the order creation time and the expected delivery time, generating the order urgency level. The smaller the difference, the higher the urgency level. The calculated transportation distance data is used as the cargo transportation distance in kilometers. The destination is encoded, specifically by using one-hot encoding to convert the destination into a numerical feature. Based on actual business needs, a logistics priority label is defined for each order. The order dataset is divided into a training set and a test set. A support vector machine model is selected and trained using the training set. Order urgency, cargo transportation distance, and destination are used as feature inputs, and priority labels are used as target outputs. Model parameters are adjusted to optimize classification accuracy. The trained support vector machine model is used to predict the test set to generate a logistics priority score for each order. The model's prediction effect is evaluated using accuracy, recall, and F1-score. The prediction results are combined with the original order data to generate a logistics priority order dataset, which contains order ID, order urgency score, cargo transportation distance, destination code, and predicted logistics priority.

[0035] The urgency of the order, the distance of the cargo transportation, and the destination are analyzed through scatter plots to view the relationship and distribution of the features and obtain visualization results. Based on the visualization results, it is determined whether the data is linearly separable. When the features show a linear separable trend, the linear SVM is selected as the support vector machine model. The linear SVM uses a linear decision boundary to divide different categories. When the features show a nonlinear distribution, the RBF kernel is selected as the support vector machine model. The RBF kernel can map the data to a high-dimensional space and find linear boundaries in the high-dimensional space.

[0036] In this example, there are 1,000 order records. Each order contains the order creation time, expected delivery time, freight transportation distance, destination information, and a logistics priority tag. The order destination is converted into a numerical feature and processed using one-hot encoding. The order's urgency score is based on the difference between the order creation time and the expected delivery time. The smaller the difference, the higher the urgency. For example, some orders with a difference of 2 days receive an urgency score of 8 (out of 10), while orders with a difference of 5 days receive an urgency score of 4. The freight transportation distance data in the orders ranges from 10 kilometers to 1,500 kilometers. For example, some orders have a transportation distance of 100 kilometers, while others have a transportation distance of 1,200 kilometers. The order destination is converted into a numerical feature using one-hot encoding. For example, for three different destinations, the encoding is: Destination A: [1, 0, 0], Destination B: [0, 1, 0], Destination C: [0, 0, 1]. The logistics priority tag is defined based on actual business needs and is categorized as "high priority" or "low priority." In the data, orders with high urgency and short transportation distance are marked as high priority, while orders with low urgency and long transportation distance are marked as low priority.

[0037] The order dataset was divided into a training set and a test set in an 8:2 ratio, with 800 orders used for model training and 200 for model testing. Scatter plots were used to analyze features such as order urgency, freight distance, and destination code to observe the relationships between these features. Experiments revealed that the order data exhibited a nonlinear distribution, so a support vector machine model with an RBF kernel was selected. The RBF kernel maps data into a high-dimensional space, enabling the identification of a linear decision boundary in that space. The SVM model was trained using the training set, using the order urgency score, freight distance, and destination code as input features and the logistics priority as the target output. Grid search was used to adjust the SVM model parameters (such as C and γ) to optimize classification accuracy.

[0038] The model used the test set to generate a logistics priority score for each order. The model's prediction performance was evaluated using precision, recall, and F1-score. The experimental results showed an accuracy of 85%, a recall of 82%, and an F1-score of 83%.

[0039] According to step S300, the logistics transportation path is optimized by using the reinforcement learning algorithm Q-learning in combination with the logistics priority of each order. The specific steps are as follows: In this example, a logistics transport environment features five primary destinations (A, B, C, D, and E) and multiple optional transport routes. Each route has different transport times and costs, which are dynamically affected by real-time traffic conditions. The logistics transport vehicle's task is to select the optimal route to deliver the goods to the destination based on the current order priority. Based on the priority data determined in previous steps, each order is assigned a high, medium, or low priority based on urgency, transport distance, and destination. There are five optional routes: P1, P2, P3, P4, and P5, each with different timeliness, transport costs, and traffic conditions. Based on real-time traffic data, traffic conditions along the transport routes are categorized as smooth, medium, and congested.

[0040] S301. Define a state space, where each state represents a specific transportation environment, including the current transportation route, cargo destination, real-time traffic conditions, transportation costs, and logistics priority of the current order. Each state is represented by a vector. Each state represents a snapshot of the logistics environment. The state vector includes: the current transportation route (P1 to P5), the destination (A to E), the real-time traffic conditions (smooth, medium, congested), the transportation cost (in RMB), and the logistics priority of the current order (high, medium, low). For example, a state can be represented as: [P1, B, medium, 100 RMB, high priority].

[0041] S302. Define the action space, where actions represent selectable transportation paths or modes of transportation. Set up a reward mechanism to encourage Q-learning to optimize the path. Select different transportation paths (P1, P2, P3, P4, P5) and different transportation modes.

[0042] S303. Create a Q table and initialize the state-action value function Q(s, a) to 0, which represents the expected reward for taking a specific action a in a specific state s. Use the ε-greedy strategy to select actions. In the early stages of training, randomly select actions to explore the environment. As learning deepens, gradually reduce the randomness and select the action with the highest current Q value. After taking an action, observe the changes in the environment, obtain the new state and reward, and update the Q value. The update formula is: ; in, is the learning rate, r is the immediate reward, is the discount factor, For the new state, is a new action; after several iterations, by continuously updating the Q value, the model's strategy for transport path selection is enhanced and the optimal path is gradually found.

[0043] Transport timeliness and cost are used as the primary reward mechanism. The higher the timeliness and the lower the cost, the higher the reward. High rewards are given for timely delivery of goods to their destination; negative rewards are imposed for delays due to poor routing. The Q(s, a) value is initially set to 0, representing the initial expected reward for selecting an action a in a specific state s.

[0044] In the early stages of training, a high ε value (e.g., 0.9) is used to randomly select transportation paths and modes for exploration to collect more data on state-action pairs. As training progresses, ε is gradually reduced (e.g., to 0.1), selecting more paths with the highest current Q values ​​to leverage the model's accumulated experience.

[0045] According to step S400, a K-means algorithm is selected as a clustering algorithm to cluster similar goods; the clustering algorithm is executed to divide the similar goods into a plurality of clusters, each cluster representing a group of similar goods that can be transported together, forming a similar goods set; during the clustering process, the number of clusters k is determined by the elbow rule; Merchant transaction records are extracted from the merchant order relationship chain. Merchants are regarded as nodes in the network, and transaction records are regarded as edges connecting nodes. The weights of the edges are weighted based on the transaction frequency to form a weighted undirected graph. The Dijkstra algorithm is used to analyze the transaction frequency and geographical location relationship between merchants, and the distance between merchants is set as the weight to optimize the transportation path. The relationships between the obtained merchants are integrated to form a merchant relationship set.

[0046] Extract information from similar goods sets and merchant relationship sets to identify merchants in the same region or logistics node and their corresponding similar goods sets; determine bundling rules, including geographic location, goods characteristics, and transportation costs; for geographic location, select merchants in the same city or logistics node for bundling; for goods characteristics, ensure that the bundled goods are of the same or compatible types; and for transportation costs, prioritize bundles with low transportation costs; Based on the bundled transport rules, a logistics bundling plan is generated, each plan including: a merchant list, a cargo list, and a transportation route; wherein the merchant list is the merchant ID and location of the merchants participating in the bundled transport, the cargo list is the same type of cargo and quantity for the bundled transport, and the transportation route is the optimal transportation route from the starting point to each destination; according to the bundling plan, the transportation tools are evaluated and batch transportation is carried out; and a transportation plan is formulated, including the departure time, route, and arrival time.

[0047] According to one embodiment of the present invention, Figure 2 As shown in the system structure diagram of an artificial intelligence-based logistics scheduling and management system, an artificial intelligence-based logistics scheduling and management system includes: Multi-source data acquisition and processing module: includes a data acquisition unit and a data pre-processing unit. The data acquisition unit collects multi-source data from various online links of international trade, including transaction orders, transaction logistics, and transaction merchant relationship chains. The data pre-processing unit cleans and formats the collected multi-source data, removes redundant information and abnormal data, and standardizes the multi-source data. Data classification and relationship chain generation module: includes: a data classification unit, a merchant order relationship chain generation unit, and a merchant similar order relationship chain generation unit; wherein the data classification unit divides the transaction order into a merchant order data set and a type order data set according to the transaction merchant and the type of goods respectively; the merchant order relationship chain generation unit generates a merchant order relationship chain based on the merchant order data set and the transaction merchant relationship chain; and the merchant similar order relationship chain generation unit generates a merchant similar order relationship chain based on the type order data set and the merchant order relationship chain; Logistics Priority Calculation and Path Optimization Module: This module includes a logistics priority calculation unit and a logistics path optimization unit. The logistics priority calculation unit uses a support vector machine algorithm to analyze the order urgency, cargo transportation distance, and destination in the merchant order dataset, calculates the logistics priority of each order, and generates a logistics priority order dataset. The logistics path optimization unit combines the logistics priority of each order and optimizes the logistics transportation path through the reinforcement learning algorithm Q-learning. It dynamically adjusts the transportation plan based on the cargo's destination, real-time traffic conditions, and logistics costs, giving priority to the delivery of high-priority orders. Logistics bundling plan generation module: includes: similar goods clustering analysis unit, merchant relationship analysis unit, logistics bundling plan generation unit and transportation tool scheduling unit; among them, the similar goods clustering analysis unit is based on the merchant similar order relationship chain, combined with the type order data set, to cluster the similar goods, identify the similar goods that can be transported together, and form a similar goods set. The merchant relationship analysis unit uses the merchant order relationship chain to identify merchants with order relationships, and analyzes the transaction frequency and geographical location relationship between merchants through the Dijkstra algorithm to form a merchant relationship set. The logistics bundling plan generation unit generates a logistics bundling plan based on the similar goods set and the merchant relationship set, and bundles merchants in the same area or logistics node with similar goods for transportation. The transportation tool scheduling unit dispatches logistics vehicles or other transportation tools according to the generated bundled transportation plan to arrange batch transportation of goods.

[0048] In this example, a logistics network covers 10 cities, with 50 merchants and 200 different types of goods. These merchants are located in different cities, have certain trading relationships, and have a certain amount of goods that need to be transported to different logistics nodes or destinations. The transaction frequency and order relationships between merchants are determined by order data. Merchants A, B, C, etc. have varying degrees of trading interactions. Goods are categorized by type, mainly food, household goods, electronic products, etc. Each type of goods has different attributes (such as volume, weight, special transportation requirements, etc.).

[0049] Using the similar goods clustering analysis unit, we clustered 200 different types of goods. Based on the merchant's similar order relationship chain and type order data set, combined with the volume, weight and transportation requirements of the goods, we identified similar goods that can be transported together and formed a similar goods set. For example: Food products were clustered into Cluster 1, which primarily included frozen foods and room-temperature foods; household goods were clustered into Cluster 2, which primarily included furniture and appliances; and electronic goods were clustered into Cluster 3, which primarily included electronic products, mobile phones, and computers. Ultimately, five clusters of similar goods were formed: food, household goods, electronic products, clothing, and other small commodities.

[0050] Using merchant order relationship chains, the Dijkstra algorithm analyzes transaction frequency and geographic location relationships between merchants. Merchants A and B frequently trade with each other and are geographically close, so the system identifies them as high-frequency traders. The Dijkstra algorithm generates merchant relationship sets based on transaction frequency, location, and order volume. For example, merchants A and B frequently trade with each other and are located close together, forming a merchant relationship set. Merchants C and D, however, have fewer transactions and are geographically distant, so they are not grouped together. This analysis yields four merchant relationship sets, each containing merchants with high transaction frequency and close geographic proximity.

[0051] Based on the set of similar goods and merchant relationships, a logistics bundling plan is generated. The logistics bundling plan generation unit identifies merchants and goods that can be bundled for transportation, bundling merchants and similar goods within the same region or logistics node for transportation. For example, merchants A and B in Cluster 1 (food) and merchants E and F in Cluster 2 (home furnishings) can be bundled together in the same region and shipped to the same logistics node. Merchants G and H in Cluster 3 (electronics) are bundled for transportation due to their high transaction frequency and geographical proximity.

[0052] The transport dispatch unit arranged appropriate transport vehicles based on the bundling plan. Two trucks were dispatched to handle the transport of similar cargo clusters: the first truck was responsible for transporting food and household items, and the second truck was responsible for transporting electronic products. The transport dispatch took into account the cargo's transportation requirements, such as cold chain transportation for food and shockproofing for electronic products.

[0053] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A logistics scheduling management method based on artificial intelligence, characterized in that: The method comprises the following steps: S100, collecting multi-source data from various online links of international trade, wherein the multi-source data includes: transaction orders, transaction logistics, and transaction merchant relationship chains; cleaning and formatting the collected multi-source data, removing redundant information and abnormal data, and standardizing the multi-source data; S200, dividing the transaction order into a merchant order data set and a type order data set according to the transaction merchant and the type of goods; generating a merchant order relationship chain based on the merchant order data set and the transaction merchant relationship chain; generating a merchant same-type order relationship chain based on the type order data set and the merchant order relationship chain; S300, using the support vector machine algorithm to analyze the order urgency, cargo transportation distance and destination in the merchant order data set, calculate the logistics priority of each order, and generate a logistics priority order data set; combining the logistics priority of each order, optimizing the logistics transportation path through the reinforcement learning algorithm Q-learning, dynamically adjusting the transportation plan according to the destination of the cargo, real-time traffic conditions and logistics costs, and giving priority to the delivery of high-priority orders; S400. Based on the merchant order relationship chain of the same type and in combination with the type order data set, cluster analysis is performed on the same type of goods to identify the same type of goods that can be transported together and form a set of the same type of goods; using the merchant order relationship chain, merchants with order relationships are identified, and the transaction frequency and geographical location relationship between merchants are analyzed through the Dijkstra algorithm to form a merchant relationship set; based on the set of the same type of goods and the merchant relationship set, a logistics bundling plan is generated to bundle merchants in the same region or logistics node with the same type of goods for transportation; according to the generated bundling transportation plan, logistics vehicles or other means of transportation are dispatched to arrange the batch transportation of goods.

2. The logistics scheduling management method based on artificial intelligence according to claim 1 is characterized by: According to step S100, the transaction orders are automatically collected from the e-commerce platform, B2B transaction website and enterprise management system, including order number, transaction time, goods details, transaction quantity, transaction price and information of both parties; the transaction logistics are obtained through the logistics management system and the third-party logistics service provider, including the delivery time, transportation status, delivery route and estimated arrival time; the contact information and transaction history of each transaction merchant are collected to form a merchant relationship chain; De-duplicate the collected multi-source data to ensure that each data record is unique; identify and eliminate outliers through Z-score analysis; process data from different sources in a unified format to ensure consistency in field names and data types; convert numerical data into numerical format, and convert categorical data into corresponding coding formats; perform multi-source data fusion and integrate the cleaned and formatted data into the data warehouse.

3. The logistics scheduling management method based on artificial intelligence according to claim 1 is characterized by: According to step S200, merchant information is extracted from the buyer and seller information in the transaction order, and the orders are grouped according to merchant ID to form a merchant order data set; Extract the cargo type information from the cargo details in the transaction order, group the orders by cargo type, and generate a type order data set; Based on the merchant order data set and combined with the transaction merchant relationship chain, a merchant order relationship chain is established; The merchant order relationship chain records the transaction relationship between different merchants and shows the mutual influence of each merchant's orders; defines the structure of the merchant order relationship chain, where nodes represent merchants, edges represent transaction relationships, and each edge is attached with information in the transaction order; Using the type order data set, we identify the orders of the same type of goods in each merchant's order data set, and determine the connections between the same type of orders by analyzing the merchant order relationship chain; we build a merchant order relationship chain of the same type to record the relationships between orders of the same type of goods and the merchants to which they belong.

4. The logistics scheduling management method based on artificial intelligence according to claim 3 is characterized by: Extract the product details of each product from the type order dataset and associate it with the merchant order dataset; Using the goods classification labels, the corresponding goods in each merchant order data set are compared with the goods in the type order data set, and the same type of goods are identified according to the preset classification standards; Extract transaction records between merchants from the merchant order data set and establish a preliminary merchant order relationship chain, with each merchant regarded as a node and transaction records as edges connecting nodes; Based on the transaction quantity and transaction price, the strength of the relationship between merchants is evaluated and represented in the form of a weighted graph. The weight of the edge represents the closeness of the transaction between merchants. The identified similar goods are combined with the merchant order relationship chain to build a merchant order relationship chain of the same type. Each edge represents an order for the same type of goods and the relationship between the merchants to which the orders belong; the node of each merchant's same type of order relationship chain represents the merchant, the edge represents the order relationship for the same type of goods, and each edge is attached with information in the transaction order.

5. The logistics scheduling management method based on artificial intelligence according to claim 1 is characterized by: According to step S300, the urgency score of the order is calculated according to the difference between the order creation time and the expected delivery time, and the urgency of the order is generated. The smaller the difference, the higher the urgency; Use the calculated transportation distance data as the cargo transportation distance in kilometers; Encode the destination, specifically using one-hot encoding to convert the destination into a numerical feature; define the logistics priority label for each order based on actual business needs; The order data set is divided into a training set and a test set. The support vector machine model is selected and the model is trained using the training set. The order urgency, cargo transportation distance and destination are used as feature inputs, and the priority label is used as the target output. The model parameters are adjusted to optimize the classification accuracy. The trained support vector machine model is used to predict the test set and generate a logistics priority score for each order. The accuracy, recall, and F1-score are used to evaluate the prediction effect of the model. The prediction results are combined with the original order data to generate a logistics priority order dataset, which includes: order ID, order urgency score, cargo transportation distance, destination code, and predicted logistics priority.

6. The method for logistics scheduling management based on artificial intelligence according to claim 5 is characterized by: The urgency of the order, the distance of cargo transportation and the destination are analyzed through a scatter plot, the relationship and distribution of the features are checked, and a visualization result is obtained. According to the visualization result, whether the data is linearly separable is determined; when the features show a linearly separable trend, a linear SVM is selected as the support vector machine model. The linear SVM uses a linear decision boundary to divide different categories; when the features show a nonlinear distribution, an RBF kernel is selected as the support vector machine model. The RBF kernel can map the data to a high-dimensional space and find linear boundaries in the high-dimensional space.

7. The method for logistics scheduling management based on artificial intelligence according to claim 6 is characterized by: According to step S300, the logistics transportation path is optimized by using the reinforcement learning algorithm Q-learning in combination with the logistics priority of each order. The specific steps are as follows: S301, define a state space, where each state represents a specific transportation environment, including the current transportation route, cargo destination, real-time traffic conditions, transportation costs, and logistics priority of the current order, and each state is represented by a vector; S302, defining an action space, where an action represents an optional transportation path or transportation mode; Set up a reward mechanism to encourage Q-learning to optimize the path; S303, create a Q table, initialize the state-action value function Q(s, a) to 0, indicating the expected reward for taking a specific action a in a specific state s; use the ε-greedy strategy to select actions. In the early stage of training, randomly select actions to explore the environment. As learning deepens, gradually reduce randomness and select the action with the highest current Q value; after taking an action, observe changes in the environment, obtain new states and rewards, and update the Q value. The update formula is: ； in, is the learning rate, r is the immediate reward, is the discount factor, For the new state, is a new action; after several iterations, by continuously updating the Q value, the model's strategy for transport path selection is enhanced and the optimal path is gradually found.

8. The logistics scheduling management method based on artificial intelligence according to claim 1 is characterized by: According to step S400, a K-means algorithm is selected as a clustering algorithm to cluster similar goods; the clustering algorithm is executed to divide similar goods into a plurality of clusters, each cluster representing a group of similar goods that can be transported together, forming a set of similar goods; During the clustering process, the number of clusters k is determined by the elbow rule; The merchant's transaction records are extracted from the merchant order relationship chain. The merchants are regarded as nodes in the network, and the transaction records are regarded as edges connecting the nodes. The weights of the edges are weighted based on the transaction frequency to form a weighted undirected graph. The Dijkstra algorithm is used to analyze the transaction frequency and geographical location relationship between merchants, and the distance between merchants is set as the weight to optimize the transportation path. Integrate the obtained relationships between merchants to form a merchant relationship set.

9. The method for logistics scheduling management based on artificial intelligence according to claim 8 is characterized by: Extract information from the same goods set and merchant relationship set to determine merchants in the same region or logistics node and their corresponding same goods set; determine the rules for bundled transportation, including: geographical location, goods characteristics and transportation cost; for geographical location, select merchants in the same city or logistics node for bundling; for goods characteristics, ensure that the bundled goods are of the same or compatible types; for transportation cost, give priority to bundle combinations with low transportation costs; Based on the bundled transport rules, a logistics bundling plan is generated, each plan includes: a merchant list, a goods list and a transportation route; wherein the merchant list is the merchant ID and location of the merchants participating in the bundled transport, the goods list is the same type of goods and quantity for bundled transport, and the transportation route is the optimal transportation route from the starting point to each destination; according to the bundling plan, the transportation tools are evaluated and batch transportation is carried out; and a transportation plan is formulated, including the departure time, route and arrival time.

10. A logistics scheduling management system based on artificial intelligence, using a logistics scheduling management method based on artificial intelligence as claimed in any one of claims 1 to 9, characterized in that: include: Multi-source data collection and processing module: including: data collection unit and data preprocessing unit; wherein the data collection unit collects multi-source data from various online links of international trade, and the multi-source data includes: transaction orders, transaction logistics and transaction merchant relationship chain; the data preprocessing unit cleans and formats the collected multi-source data, removes redundant information and abnormal data, and standardizes the multi-source data; Data classification and relationship chain generation module: including: data classification unit, merchant order relationship chain generation unit and merchant same-type order relationship chain generation unit; wherein the data classification unit is based on the transaction order and is divided into a merchant order data set and a type order data set according to the transaction merchant and the type of goods, the merchant order relationship chain generation unit generates a merchant order relationship chain based on the merchant order data set and the transaction merchant relationship chain, and the merchant same-type order relationship chain generation unit generates a merchant same-type order relationship chain based on the type order data set and the merchant order relationship chain; Logistics priority calculation and path optimization module: including: logistics priority calculation unit and logistics path optimization unit; the logistics priority calculation unit uses the support vector machine algorithm to analyze the order urgency, cargo transportation distance and destination in the merchant order data set, calculates the logistics priority of each order, and generates a logistics priority order data set. The logistics path optimization unit combines the logistics priority of each order and optimizes the logistics transportation path through the reinforcement learning algorithm Q-learning, dynamically adjusts the transportation plan according to the destination of the cargo, real-time traffic conditions and logistics costs, and gives priority to the delivery of high-priority orders; The logistics bundling scheme generation module includes: a clustering analysis unit for similar goods, a merchant relationship analysis unit, a logistics bundling scheme generation unit and a transport tool scheduling unit; wherein the clustering analysis unit for similar goods performs clustering analysis on similar goods based on merchant similar order relationship chains and in combination with type order data sets, identifies similar goods that can be transported together, and forms a similar goods set; the merchant relationship analysis unit uses merchant order relationship chains to identify merchants with order relationships, and analyzes transaction frequencies and geographical location relationships between merchants through the Dijkstra algorithm to form a merchant relationship set; the logistics bundling scheme generation unit generates a logistics bundling scheme based on the similar goods set and merchant relationship set, and bundles merchants in the same region or logistics node with similar goods for transportation; the transport tool scheduling unit dispatches logistics vehicles or other transportation tools according to the generated bundling transportation scheme, and arranges batch transportation of goods.

Images