Material order intelligent management method and management system

Abstract

The invention relates to the technical field of order management, in particular to an intelligent material order management method and system, and the method comprises the following steps: based on collected order data, analyzing the feature expression of the order data through an auto-encoder, comparing the feature data with newly input order data, and calculating a difference value; and marking the order with the difference value exceeding a threshold value to generate an abnormal order identifier. According to the invention, the self-encoder is used to deeply analyze order data, automatically identify orders having significant differences with normal data, and carry out abnormal marking in real time, so that execution errors are effectively prevented and fraud risks are reduced, and the application of the long and short term memory network further enhances the prediction capability of abnormal order behaviors. Through the analysis of the time sequence data, the potential order execution problem can be tracked and predicted, the constructed dynamic relation graph between the materials and the orders provides a global view for the material flow and inventory management, and the optimization of the material use and the cost control are promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of order management, and in particular to an intelligent material order management method and management system. Background Art

[0002] Order management technology mainly focuses on methods to effectively process and track orders in business operations, involving all aspects from order generation, inventory management, logistics distribution to customer satisfaction evaluation. With the development of information technology, modern order management increasingly relies on automated tools to improve efficiency and accuracy, and realizes real-time data processing and updating by integrating data from sales, inventory and customer relations.

[0003] Among them, the intelligent management method of material orders is a management system that uses information technology to optimize material procurement, storage, use and replenishment. It is mainly used in manufacturing and supply chain management. The purpose is to accurately predict material requirements through intelligent systems, ensure the continuous operation of production lines and maximize cost efficiency, so as to achieve real-time monitoring and processing of orders, thereby reducing inventory backlogs and improving the ability to respond to market changes.

[0004] Existing technologies are not sufficient to capture immediate changes in order execution in static data processing, resulting in response delays and inefficient processing. The lack of effective anomaly detection and prediction mechanisms often leads to the accumulation of problems, which in turn affects customer satisfaction. In terms of material management, traditional methods cannot accurately predict material requirements, resulting in inventory backlogs or shortages, which not only increases costs but also interrupts production due to insufficient materials. Logistics management lacks dynamic optimization capabilities and cannot adjust logistics plans in real time to respond to emergencies in the supply chain, thereby increasing transportation costs and time. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent material order management method and management system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a material order intelligent management method, comprising the following steps:

[0007] S1: Based on the collected order data, the autoencoder is used to analyze the feature expression of the order data, compare the feature data with the newly input order data, calculate the difference value, and mark the orders whose difference value exceeds the threshold to generate an abnormal order mark;

[0008] S2: Based on the abnormal order identification, the LSTM network is used to analyze the time series data of the abnormal orders, identify the abnormal orders with large deviations from the normal order time flow, and obtain abnormal time identification information;

[0009] S3: Based on the abnormal time identification information, a dynamic relationship graph between materials and orders is constructed, graph data changes are monitored in real time, the interaction behaviors between material nodes in the graph are analyzed, and the correlation and dependency between materials are identified to generate a material dependency graph;

[0010] S4: Based on the material dependency graph, the Dijkstra algorithm is used to calculate the optimal path from the supply node to the demand node, identify the key connection points in the path, and optimize the material circulation efficiency and order process to obtain the order management optimization configuration.

[0011] The present invention is improved in that the analysis steps of the characteristic expression of the order data are specifically as follows:

[0012] S111: extract key information based on the collected order data, use the autoencoder to encode the extracted data, and obtain the encoded data features;

[0013] S112: Execute a clustering algorithm on the encoded data features, using the formula:

[0014]

[0015] Calculate the distance D(c i ,x j ), and obtain the order data feature expression, where c i represents the i-th cluster center, x j Represents the feature vector of the j-th data point after encoding.

[0016] The present invention is improved in that the step of obtaining the abnormal order identifier is specifically as follows:

[0017] S121: Compare the feature data with the newly input order data, iterate each new order data, determine the distance from the cluster center, and obtain a difference value array;

[0018] S122: Based on the difference value array, adopt the formula:

[0019] δ Q =μ Q +k Q ×σ Q

[0020] Define the anomaly threshold δ Q , and mark the orders whose difference value exceeds the threshold, and generate abnormal order identification, where μ Q is the mean of the difference array, σ Q is the standard deviation of the difference array, k Q It is a coefficient adjusted according to the frequency of historical abnormal orders.

[0021] The present invention is improved in that the analysis steps of the time series data are specifically as follows:

[0022] S211: Based on the abnormal order identifier, key timestamp data of the order is collected, including the submission, processing and completion time of the order, and the data is sorted to obtain a time series data set;

[0023] S212: Analyze the time series data set using the LSTM network, using the formula:

[0024]

[0025] Get the abnormal score AS at the time point t , where TS t represents the order timestamp at the target time point t, is the average value of the timestamp, SR t Indicates the standard deviation of timestamps.

[0026] The present invention is improved in that the step of obtaining the abnormal time identification information is specifically as follows:

[0027] S221: Capture abnormal points that deviate from the normal order processing time flow and obtain a list of abnormal time points;

[0028] S222: Performing behavior analysis on the abnormal time point list, using the formula:

[0029] VScore=FD t -μ FD

[0030] Compare the deviation value VScore between the order duration at each time point and the average duration to obtain abnormal time identification information, where FD t is the order duration at time t, μ FD is the average of the order durations.

[0031] The present invention is improved in that the step of real-time monitoring of graph data changes is specifically as follows:

[0032] S311: Based on the abnormal time identification information, extract material and order data, including interaction time, material number and order number, and construct material order basic data;

[0033] S312: Based on the material order basic data, adjust the relationship strength using the formula:

[0034]

[0035] Real-time update of dynamic relationship diagram, where RS ij represents the strength of the relationship between nodes i and j, FSij is the interaction frequency between material i and order j, α Y and β Y is the adjustment coefficient, k Y is the curvature adjustment parameter of the Sigmoid function, τ Y is the interaction frequency threshold;

[0036] S313: Monitor the data changes of the dynamic relationship diagram, analyze the interaction behaviors between material nodes in the diagram, and obtain the material and order interaction analysis results.

[0037] The present invention is improved in that the step of obtaining the material dependency graph is specifically as follows:

[0038] S321: Collect the interaction data between materials and orders, and calculate the interaction intensity between each pair of materials using the formula:

[0039]

[0040] Get material dependency data, where RB ij represents the dependence strength of material i on j, wb ij represents the interaction strength between materials i and j, γ b is the adjustment coefficient, ∑ k wb ik It is the sum of the interaction intensities between material i and other materials;

[0041] S322: Based on the material dependency data, identify the dependency structure and association network between materials to obtain a material dependency graph.

[0042] The present invention is improved in that the steps of obtaining the order management optimization configuration are specifically as follows:

[0043] S411: Based on the material dependency graph, applying the Dijkstra algorithm, according to the current supply chain network data, calculating the optimal path from the supply node to the demand node, and generating optimal path data;

[0044] S412: Based on the optimal path data, the key connection point index of each node is calculated using the formula:

[0045]

[0046] The node criticality score KI is obtained, where TD represents the total accumulated delay time of the node, and TG represents the total number of transactions passing through the node;

[0047] S413: Based on the node criticality scores, reconfigure the material flow path and order processing flow, optimize the supply chain efficiency, and obtain an optimized order management configuration.

[0048] A material order intelligent management system, the system comprising:

[0049] The order feature analysis module uses the autoencoder to perform difference analysis with the new order data based on the collected order data, compares the feature data with the newly input order data, and marks the orders whose difference values ​​exceed the threshold to generate abnormal order identification;

[0050] The abnormal order identification module uses an LSTM network to analyze the time series data of abnormal orders according to the abnormal order identification, identifies abnormal orders with large deviations from normal order time processes, and obtains abnormal time identification information;

[0051] The material relationship diagram module establishes a dynamic relationship diagram between materials and orders based on the abnormal time identification information, monitors and analyzes the interaction between material nodes in real time, identifies the correlation and dependency between materials, and generates a material dependency diagram;

[0052] The path optimization module uses the Dijkstra algorithm based on the material dependency graph to determine the optimal path from the supply node to the demand node, and obtains the optimal logistics path through path calculation and key node identification;

[0053] The order process optimization module performs order circulation and process optimization according to the optimal logistics path, optimizes order processing efficiency through path analysis and material circulation adjustment, and obtains order management optimization configuration.

[0054] Compared with the prior art, the advantages and positive effects of the present invention are:

[0055] In the present invention, an autoencoder is used to conduct in-depth analysis of order data, automatically identify orders that are significantly different from normal data, and mark anomalies in real time, thereby effectively preventing execution errors and reducing fraud risks. The application of long short-term memory networks further enhances the ability to predict abnormal order behaviors. Through the analysis of time series data, potential order execution problems can be tracked and foreseen. The constructed dynamic relationship diagram between materials and orders provides a global view for material flow and inventory management, promotes the optimization of material use and cost control, and optimizes the material flow and refines the overall operation of the supply chain to improve management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A flow chart of a material order intelligent management method is proposed for the present invention;

[0057] Figure 2 This is a flow chart for analyzing the feature expression of order data in the present invention;

[0058] Figure 3 This is a flow chart for obtaining abnormal order identification in the present invention;

[0059] Figure 4 It is a flow chart of the analysis of time series data in the present invention;

[0060] Figure 5 This is a flow chart for obtaining abnormal time identification information in the present invention;

[0061] Figure 6 A flowchart of real-time monitoring of graph data changes in the present invention;

[0062] Figure 7 It is a flow chart for obtaining the material dependency graph in the present invention;

[0063] Figure 8 This is a flowchart for obtaining the order management optimization configuration in the present invention. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0065] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0066] Example

[0067] See also Figure 1 The present invention provides a technical solution: a material order intelligent management method, comprising the following steps:

[0068] S1: Based on the collected order data, the autoencoder is used to analyze the feature expression of the order data, compare the feature data with the newly input order data, calculate the difference value, and mark the orders whose difference value exceeds the threshold to generate an abnormal order mark;

[0069] S2: Based on the abnormal order identification, the LSTM network is used to analyze the time series data of abnormal orders, identify abnormal orders with large deviations from the normal order time flow, and obtain abnormal time identification information;

[0070] S3: Based on the abnormal time identification information, a dynamic relationship graph between materials and orders is constructed, graph data changes are monitored in real time, the interaction behaviors between material nodes in the graph are analyzed, and the correlation and dependency between materials are identified to generate a material dependency graph;

[0071] S4: Based on the material dependency graph, the Dijkstra algorithm is used to calculate the optimal path from the supply node to the demand node, identify the key connection points in the path, optimize the material flow efficiency and order process, and obtain the optimal configuration of order management.

[0072] Abnormal order identification includes order deviation rate, marking time point, and marking reason. Abnormal time identification information includes deviation duration, deviation start time, and deviation impact range. The material dependency graph includes node interaction type, node connection strength, and node activity. The order management optimization configuration includes the adjusted circulation path and key node list.

[0073] See also Figure 2 ,The analysis steps of the feature expression of order data are as follows:

[0074] S111: Based on the collected order data, extract key information, including product category, transaction amount, customer rating and purchase frequency, and use the autoencoder to encode the extracted data to obtain the encoded data features;

[0075] By sorting out historical order data, key information such as product category, transaction amount, customer rating and purchase frequency is extracted. This information is used as input features of the autoencoder. By learning the features, the autoencoder can extract the complex patterns and deep relationships hidden in the raw data. The process requires establishing an appropriate neural network structure and adjusting the network parameters to optimize the encoding expression of the features. The generated encoded data feature expression is the input of the next clustering algorithm.

[0076] S112: Execute a clustering algorithm on the encoded data features using the formula:

[0077]

[0078] Calculate the distance D(c i ,x j ), and obtain the order data feature expression, where c i represents the i-th cluster center. The cluster center is determined by the clustering algorithm (K-means), which is determined by continuously optimizing the sum of the squares of the distances between the data points and the center points during the iteration process. j The feature vector representing the j-th data point after encoding. The feature vector is obtained from data preprocessing and feature extraction. For example, the high-dimensional data extracted from the original order data by the autoencoder is used as clustering input after dimensionality reduction.

[0079] The following data is collected, assuming c i =(1,2) as the cluster center, x j =(3,5) as a data point, substitute it into the formula to calculate:

[0080]

[0081] The distance between this data point and the cluster center is approximately 3.605. The distance value will be used to determine the category of the data point. The class center with the smallest distance will be the label of the point, ensuring that each point is assigned to the nearest cluster center.

[0082] See also Figure 3 , the specific steps for obtaining abnormal order identification are:

[0083] S121: Compare the feature data with the newly input order data, iterate each new order data, determine the distance from the cluster center, and obtain a difference value array;

[0084] Calculate the feature difference between each new order and the existing category center by extracting the feature vector of each order, such as product type, purchase quantity, price, etc. Specifically, for example, if an order contains 10 items A with a unit price of 20, its feature vector is represented as [product A, 10, 20], and perform Euclidean distance calculation on the feature vector and the feature vector of each category center obtained during the cluster analysis. The calculation of Euclidean distance is based on the shortest straight line distance between two points, and the calculation formula is:

[0085]

[0086] in, and are two feature vectors, p1 and q1 are the numerical values ​​of the corresponding features. In this way, each new order will get an array of difference values ​​from the category center. The difference values ​​constitute an array for subsequent abnormal order detection. Abnormal orders are marked by comparing whether the difference values ​​exceed the threshold.

[0087] S122: Based on the difference value array, use the formula:

[0088] δ Q =μ Q +k Q ×σ Q

[0089] Define the anomaly threshold δ Q , and mark the orders whose difference value exceeds the threshold, and generate abnormal order identification, where μ Qis the average value of the difference value array, taken from the difference values ​​of all orders in the historical order data, and the calculation formula is σ Q is the standard deviation of the difference value array, which is based on the difference value of the same batch of historical order data, and the calculation formula is k Q k is a coefficient adjusted according to the frequency of historical abnormal orders. It is based on the needs of risk management and is set according to the proportion of abnormal orders in historical data and the business's tolerance for risk. It can be quantified by analyzing the ratio of abnormal orders to normal orders in historical data and the business impact of abnormal events. Q For example, if historical data shows that a higher frequency of abnormal orders is associated with significant business disruptions, k Q Needs to be set higher to more conservatively flag potential anomalies;

[0090] Calculate the mean μ of the difference value array from historical data Q =15 and standard deviation σ Q =5, set k Q =2, then the threshold is calculated as:

[0091] δ Q =15+2×5=25

[0092] This means that any order with a difference value exceeding 25 will be marked as abnormal. The set threshold can effectively identify abnormal situations from new orders. Orders with difference values ​​significantly higher than normal levels will be marked and subsequently processed to ensure the normal operation of order processing and risk control.

[0093] See also Figure 4 , the analysis steps of time series data are as follows:

[0094] S211: Based on the abnormal order identification, key timestamp data of the order is collected, including the submission, processing and completion time of the order, and the data is sorted to obtain a time series data set;

[0095] Based on the existing abnormal order identification, each order record marked as abnormal is retrieved, involving specific order processing key timestamps, including but not limited to the order submission time, processing start and end time. The data is organized and serialized to construct a time series data set. This data set will be directly used for subsequent deep learning processing and analysis to ensure that all relevant data is systematically collected and organized, providing data support for accurate analysis.

[0096] S212: Use LSTM network to analyze the time series data set, using the formula:

[0097]

[0098] Get the abnormal score AS at the time point t , where TS t The order timestamp representing the target time point t is the timestamp data obtained directly from the real-time order processing system. is the average of the timestamps, based on the timestamps of all orders in the order data. t represents the standard deviation of timestamps, also based on order data, γ S and δ S is a model parameter, adjusted to optimize the accuracy of anomaly identification, γ S Used to adjust the impact of the difference between the timestamp and the average value on the anomaly score, δ S It is the effect of adjusting the standard deviation on the anomaly score, which is optimized by the gradient descent method. The goal is to minimize the error of the model on the validation dataset;

[0099] If the order timestamp TS at time point t t = 150 minutes, average value of timestamp Minutes, standard deviation SR of timestamp t = 5 minutes, model parameter γ S =1.2,δ S =0.8, then the anomaly score is calculated as follows:

[0100] AS t =1.2·(150-145)+0.8·5=6+4=10

[0101] The result shows that at time point t, the order timestamp is higher than the average time, and after adjustment for the standard deviation, the score is 10, which means that according to the set threshold (if it is 5, more than 5 is abnormal), this time point is identified as an anomaly.

[0102] See also Figure 5 ,The specific steps for obtaining abnormal time identification information are:

[0103] S221: Capture abnormal points that deviate from the normal order processing time flow and obtain a list of abnormal time points;

[0104] Key time markers are extracted from multiple order data streams, such as the timestamps of order start and completion. Data points are considered abnormal points and require further analysis and confirmation to ensure that potential abnormal orders that deviate from the normal process are quickly captured from a large amount of data, and a list of time points containing suspected abnormal orders is generated. This list provides basic data for subsequent behavioral analysis, making problem location more accurate.

[0105] S222: Perform behavioral analysis on the abnormal time point list, using the formula:

[0106] VScore=FD t -μ FD

[0107] Compare the deviation value VScore between the order duration at each time point and the average duration to obtain abnormal time identification information, where FD t is the order duration at time point t, and the actual completion time of each order is obtained directly from the order processing system, μ FD It is the average value of order duration. It is based on the duration of all completed orders in the order data. By measuring the actual completion time of each order and calculating the average time, the order processing efficiency can be effectively monitored and analyzed. It aims to detect abnormal fluctuations in order processing time. Abnormal fluctuations include process interruptions, resource allocation problems, or the existence of non-standard operations.

[0108] The order duration FD at time point t t = 300 minutes, and the average order duration μ FD =200 minutes, then the difference calculation process is:

[0109] VScore=300-200=100

[0110] The result shows that at time point t, the duration of the order is 100 minutes longer than the average time, and the difference is far beyond the normal range, indicating that it is an abnormal order.

[0111] See also Figure 6 ,The steps for real-time monitoring of graph data changes are as follows:

[0112] S311: Based on the abnormal time identification information, extract material and order data, including interaction time, material number and order number, and construct material order basic data;

[0113] First, material-related information is extracted from the collected order data, including key data fields such as material number, interaction time and order number. The interaction time records the time series information between materials and orders. By analyzing the abnormal time identification information, those orders and material interaction data with large time process deviations can be screened out. Next, the information of relevant materials in each order is extracted to form a material and order data set with a complete time series, which includes the number of each material in the order and the corresponding interaction time. The data will be organized into a standardized basic data format for subsequent use. By extracting and organizing the data, a basic material order data set based on factors such as time and number can be formed, providing basic data support for subsequent material dependency analysis and order management optimization.

[0114] S312: Based on the basic data of material orders, adjust the relationship strength using the formula:

[0115]

[0116] Update the dynamic relationship diagram in real time to ensure that the data in the diagram reflects the latest status. ij represents the strength of the relationship between nodes i and j, FS ij is the interaction frequency between material i and order j, obtained from the order processing system and calculated by counting the number of interactions between each material and order, α Y and β Y k is the adjustment coefficient used to balance the relationship between frequency and intensity. Based on historical data analysis, the coefficient is determined to maximize the accuracy and sensitivity of the relationship graph, and then adjusted on the validation data set by gradient descent. Y is the curvature adjustment parameter of the Sigmoid function, which is set based on the distribution of interaction frequencies, τ Y The interaction frequency threshold is set based on the average interaction frequency or median in historical data, and is intended to distinguish between normal and abnormal interaction frequencies;

[0117] If we have the following data, the interaction frequency FS between node i and node j ij =5, interaction frequency threshold τ Y =3, coefficient α Y =0.5,β Y =0.8, curvature adjustment parameter k Y =1, the calculation process is:

[0118]

[0119] The result shows that the relationship strength between node i and node j is 3.204, reflecting a strong interaction frequency between the two.

[0120] S313: Monitor the data changes of the dynamic relationship graph, analyze the interaction behaviors between material nodes in the graph, and obtain the analysis results of the interaction between materials and orders;

[0121] First, the data of each node in the material dependency graph is obtained in real time, including the material number, status information, and real-time interaction of orders, etc., and the changes of these data are continuously tracked to analyze the interaction behavior between material nodes in the graph, including the material flow between nodes, the time nodes of material demand and supply, and the dependency between materials. By monitoring the interaction behavior, the interaction mode and association relationship between materials and orders in the graph are extracted. Combined with the change information of the nodes in the graph, the relationship structure changes between materials can be accurately identified, thereby obtaining the interaction analysis results between materials and orders. This analysis process continuously updates the graph data, evaluates the impact of changes between material nodes on order flow, and finally obtains the analysis results reflecting the interaction behavior between materials and orders for reference in subsequent order optimization and supply chain decision-making.

[0122] See also Figure 7 , the specific steps for obtaining the material dependency graph are:

[0123] S321: Collect the interaction data between materials and orders, and calculate the interaction intensity between each pair of materials using the formula:

[0124]

[0125] Get material dependency data, where RB ij represents the dependency strength of item i on item j, which is calculated by comparing the interaction strength of a specific item pair with the sum of the interaction strengths of that item with all other items. ij Represents the interaction intensity between materials i and j, which is obtained from material management data and calculated by counting the order frequency or usage of material i for material j. b is the adjustment coefficient, which is used to control the sensitivity of the dependency intensity. It is adjusted according to business needs and historical data analysis to ensure that the calculation of the dependency reflects the actual business impact. k wb ik It is the sum of the interaction intensities between material i and other materials, and all interaction records about material i are counted from the material management data;

[0126] If the following data are available, the interaction strength wb between materials i and j ij =50, and the sum of the interaction strengths between material i and all other materials ∑ k wb ik =200, adjustment factor γ b =0.5, then substitute into the formula to calculate:

[0127]

[0128] The result shows that the dependence intensity of material i on material j is 0.125, reflecting that the dependence between material i and material j is relatively low. Further analysis is needed to decide whether it is necessary to enhance this dependence or seek other supply options to optimize the stability and efficiency of the supply chain.

[0129] S322: Based on the material dependency data, identify the dependency structure and association network between materials to obtain a material dependency graph;

[0130] First, collect and organize the dependencies between materials. Material dependency data is usually obtained by counting historical orders, production plans, and material flow records. The dependency between each material and other materials is quantified through factors such as the co-occurrence frequency of materials, usage order, and flow path. The dependency value reflects the strength of the relationship between materials. Next, based on the data, construct an association matrix between materials, in which each matrix element represents the dependency strength between material i and material j. Larger values ​​in the matrix indicate that the dependency relationship between materials is stronger. Through cluster analysis or graph theory algorithms, further identify the dependency structure and association network between materials. The dependency structure can be represented by the edges of the graph, and the weight of the edge represents the dependency strength between materials. Finally, using the graph generation method, each material and other interdependent materials are connected through nodes and edges to obtain a complete material dependency graph, in which the nodes in the graph represent materials and the edges represent the dependency relationship between materials, providing support for subsequent supply chain optimization and order management.

[0131] See also Figure 8 , the specific steps for obtaining the order management optimization configuration are:

[0132] S411: Based on the material dependency graph, the Dijkstra algorithm is applied to calculate the optimal path from the supply node to the demand node according to the current supply chain network data, and the optimal path data is generated;

[0133] Applying the Dijkstra algorithm, we first start from each supply node in the supply chain network, and calculate the shortest path of the network nodes to obtain the distance from each supply node to the demand node. The path between the nodes is represented by the edge weight, which reflects factors such as the transportation cost or transportation time of the material. At this time, according to the network topology and the actual situation of material flow, the optimal path from the current supply node to the demand node is calculated. When optimizing the path, the total cost or time of the path is taken into account to determine the shortest distance or the lowest cost path on the path. By dynamically updating the supply chain network data, the edge weight value of the path is adjusted in real time during each path calculation to ensure that the optimal path is updated over time. The calculation result is the optimal path data between each supply node and the demand node, and the actual feasibility of the path is ensured. The calculation process is based on the collected supply chain data and is continuously iterated and updated, and finally outputs the specific information of the optimal path, including the node sequence of the path and the specific transportation cost or time of each node.

[0134] S412: Based on the optimal path data, the key connection point index of each node is calculated using the formula:

[0135]

[0136] The node criticality score KI is obtained, where TD represents the total accumulated delay time of the node, including all delays caused by material shortages, processing errors or other logistics problems, which is collected from logistics management data and calculated by counting all delay times caused by the above reasons for each node in a specific period. TG represents the total number of transactions passing through the node, reflecting the activity of the node and the size of the logistics circulation, and all transaction activities recorded in the logistics system passing through a specific node;

[0137] If the TD of a node is 50 hours and the TG is 200 transactions, then the KI is calculated as:

[0138]

[0139] The result shows that each transaction on this node is delayed by an average of 0.25 hours. The score helps identify key nodes that need to be optimized.

[0140] S413: Based on the node criticality score, reconfigure the material flow path and order processing flow to optimize the supply chain efficiency, thereby reducing costs and improving response speed, and obtaining the optimal configuration of order management;

[0141] First, score each node in the material dependency graph to evaluate the importance of each node in the overall supply chain. The criticality score is calculated based on the relative position of the node, the material flow volume, and the closeness of the dependency relationship. The score reflects the influence and criticality of the node in the entire supply chain network. Then, the material flow path is reconfigured according to the criticality score of the node, and the priority of the material flow path is adjusted according to the nodes with high scores. The paths of low-scoring nodes can be optimized or replaced. The adjusted material flow path and order processing flow will reduce unnecessary intermediate links, improve the flexibility and responsiveness of the supply chain, and ensure that the material flow of important nodes is not affected. At the same time, the order processing process is optimized to ensure that orders at key nodes are given priority, thereby reducing the bottleneck problem of the overall supply chain. Finally, the order management optimization configuration is obtained, and the configuration is continuously adjusted through real-time monitoring and feedback mechanisms to cope with supply chain changes.

[0142] A material order intelligent management system, the system comprising:

[0143] The order feature analysis module uses the autoencoder to perform difference analysis with the new order data based on the collected order data, compares the feature data with the newly input order data, and marks the orders whose difference values ​​exceed the threshold to generate abnormal order identification;

[0144] The abnormal order identification module uses LSTM network to analyze the time series data of abnormal orders according to the abnormal order identification, identify abnormal orders with large deviations from the normal order time flow, and obtain abnormal time identification information;

[0145] The material relationship diagram module establishes a dynamic relationship diagram between materials and orders based on abnormal time identification information, monitors and analyzes the interaction between material nodes in real time, identifies the correlation and dependency between materials, and generates a material dependency diagram;

[0146] The path optimization module uses the Dijkstra algorithm based on the material dependency graph to determine the optimal path from the supply node to the demand node, and obtains the optimal logistics path through path calculation and key node identification;

[0147] The order process optimization module performs order flow and process optimization according to the optimal logistics path. Through path analysis and material flow adjustment, it optimizes order processing efficiency and obtains the optimal configuration of order management.

[0148] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A material order intelligent management method, characterized in that: The following steps are involved: Based on the collected order data, the autoencoder is used to analyze the feature expression of the order data, compare the feature data with the newly input order data, calculate the difference value, and mark the orders whose difference values ​​exceed the threshold to generate abnormal order identification; Based on the abnormal order identification, an LSTM network is used to analyze the time series data of abnormal orders, identify abnormal orders with large deviations from normal order time processes, and obtain abnormal time identification information; Based on the abnormal time identification information, a dynamic relationship diagram between materials and orders is constructed, the graph data changes are monitored in real time, the interaction behaviors between material nodes in the graph are analyzed, and the correlation and dependency between materials are identified to generate a material dependency graph; Based on the material dependency graph, the Dijkstra algorithm is used to calculate the optimal path from the supply node to the demand node, identify the key connection points in the path, optimize the material circulation efficiency and order process, and obtain the order management optimization configuration.

2. The material order intelligent management method according to claim 1 is characterized in that: The analysis steps of the characteristic expression of the order data are specifically as follows: Based on the collected order data, key information is extracted, and the extracted data is encoded using an autoencoder to obtain the encoded data features; A clustering algorithm is performed on the encoded data features, using the formula: Calculate the distance D(c i ,x j ), and obtain the order data feature expression, where c i represents the i-th cluster center, x j Represents the feature vector of the j-th data point after encoding.

3. The intelligent material order management method according to claim 1, characterized in that: The specific steps for obtaining the abnormal order identifier are: Compare the feature data with the newly input order data, iterate each new order data, determine the distance from the cluster center, and obtain a difference value array; Based on the difference value array, the formula is used: d Q =μ Q +k Q ×s Q Define the anomaly threshold δ Q , and mark the orders whose difference value exceeds the threshold, and generate abnormal order identification, where μ Q is the mean of the difference array, σ Q is the standard deviation of the difference array, k Q It is a coefficient adjusted according to the frequency of historical abnormal orders.

4. The material order intelligent management method according to claim 1, characterized in that: The analysis steps of the time series data are specifically as follows: Based on the abnormal order identifier, key timestamp data of the order is collected, including the submission, processing and completion time of the order, and the data is sorted to obtain a time series data set; The time series data set is analyzed using the LSTM network using the formula: Get the abnormal score AS at the time point t , where TS t represents the order timestamp at the target time point t, is the average value of the timestamp, SR t Indicates the standard deviation of timestamps.

5. The intelligent material order management method according to claim 1, characterized in that: The steps for obtaining the abnormal time identification information are specifically as follows: Capture abnormal points that deviate from the normal order processing time flow and obtain a list of abnormal time points; The behavior analysis is performed on the abnormal time point list using the formula: VScore=FD t -μ FD Compare the deviation value VScore between the order duration at each time point and the average duration to obtain abnormal time identification information, where FD t is the order duration at time t, μ FD is the average of the order durations.

6. The material order intelligent management method according to claim 1, characterized in that: The steps of real-time monitoring of graph data changes are specifically as follows: Based on the abnormal time identification information, extract material and order data, including interaction time, material number and order number, and construct material order basic data; Based on the basic data of the material order, the relationship strength is adjusted using the formula: Real-time update of dynamic relationship diagram, where RS ij represents the strength of the relationship between nodes i and j, FS ij is the interaction frequency between material i and order j, α Y and β Y is the adjustment coefficient, k Y is the curvature adjustment parameter of the Sigmoid function, τ Y is the interaction frequency threshold; Monitor the data changes of the dynamic relationship diagram, analyze the interaction behaviors between material nodes in the diagram, and obtain the analysis results of the interaction between materials and orders.

7. The material order intelligent management method according to claim 1, characterized in that: The steps for obtaining the material dependency graph are specifically as follows: Collect the interaction data between materials and orders, and calculate the interaction intensity between each pair of materials using the formula: Get material dependency data, where RB ij represents the dependence strength of material i on j, wb ij represents the interaction strength between materials i and j, γ b is the adjustment coefficient, ∑ k wb ik It is the sum of the interaction intensities between material i and other materials; Based on the material dependency data, the dependency structure and association network between materials are identified to obtain a material dependency graph.

8. The material order intelligent management method according to claim 1, characterized in that: The steps for obtaining the order management optimization configuration are specifically as follows: Based on the material dependency graph, the Dijkstra algorithm is applied to calculate the optimal path from the supply node to the demand node according to the current supply chain network data, and the optimal path data is generated; Based on the optimal path data, the key connection point index of each node is calculated using the formula: The node criticality score KI is obtained, where TD represents the total accumulated delay time of the node, and TG represents the total number of transactions passing through the node; Based on the node criticality scores, the material flow path and order processing flow are reconfigured to optimize the supply chain efficiency and obtain an optimized order management configuration.

9. A material order intelligent management system, characterized in that: According to any one of claims 1 to 8, the material order intelligent management method is implemented, and the system comprises: The order feature analysis module uses the autoencoder to perform difference analysis with the new order data based on the collected order data, compares the feature data with the newly input order data, and marks the orders whose difference values ​​exceed the threshold to generate abnormal order identification; The abnormal order identification module uses an LSTM network to analyze the time series data of abnormal orders according to the abnormal order identification, identifies abnormal orders with large deviations from normal order time processes, and obtains abnormal time identification information; The material relationship diagram module establishes a dynamic relationship diagram between materials and orders based on the abnormal time identification information, monitors and analyzes the interaction between material nodes in real time, identifies the correlation and dependency between materials, and generates a material dependency diagram; The path optimization module uses the Dijkstra algorithm based on the material dependency graph to determine the optimal path from the supply node to the demand node, and obtains the optimal logistics path through path calculation and key node identification; The order process optimization module performs order circulation and process optimization according to the optimal logistics path, optimizes order processing efficiency through path analysis and material circulation adjustment, and obtains order management optimization configuration.

Images