User demand forecasting method and system based on supply chain settlement service system

Abstract

The present application provides a user demand prediction method and system based on a supply chain settlement service system. By performing abnormal behavior pattern analysis on the first supply chain conversation flow data monitored by the data monitoring port, the abnormal behavior pattern data of the data monitoring port can be detected in a timely and accurate manner. Secondly, the data window shielding process is executed to effectively shield the abnormal behavior pattern data to prevent it from interfering with normal user demand prediction, further ensuring the reliability of the prediction results. Thirdly, by performing abnormal behavior pattern analysis on the second supply chain conversation flow data, the execution effect of the data window shielding process can be verified to ensure that the abnormal behavior data has been completely shielded, providing a clean and accurate data basis for subsequent user demand prediction. Finally, user demand prediction based on key supply chain conversation flow data can more accurately predict market dynamics and user needs.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a method and system for predicting user demand based on a supply chain settlement service system. Background Art

[0002] With the continuous development and improvement of supply chain management technology, supply chain settlement service systems have become an indispensable part of modern enterprises. These systems usually involve a large amount of data processing and information exchange to meet the ever-changing market needs. However, when processing supply chain conversation flow data, how to accurately predict user needs and respond quickly has always been a challenge facing the industry.

[0003] Traditional user demand forecasting methods mainly rely on historical sales data, market research and other relevant statistical information. However, these methods often have limitations when processing real-time data because they cannot capture market dynamics and changes in consumer behavior in a timely manner. In addition, traditional forecasting methods are also susceptible to data anomalies and noise, resulting in inaccurate forecast results.

[0004] In recent years, with the development of big data and machine learning technology, user demand forecasting methods based on real-time data have received widespread attention. These methods can more accurately capture market dynamics and consumer preferences by analyzing supply chain conversation flow data. However, due to the complexity and diversity of supply chain data, how to effectively process and analyze this data remains a challenging problem.

[0005] In particular, traditional forecasting methods often fail to make accurate judgments when faced with abnormal behavior pattern data. These data may be generated due to system failures, malicious attacks, or abnormal market fluctuations. If they are not identified and processed in a timely manner, they will have a serious impact on the forecast results.

[0006] Therefore, it is particularly important to develop a method that can process supply chain conversation flow data in real time, accurately identify abnormal behavior patterns, and predict user demand. Summary of the invention

[0007] In view of the above-mentioned problems, in combination with the first aspect of the present application, an embodiment of the present application provides a user demand prediction method based on a supply chain settlement service system, the method comprising:

[0008] Acquire first supply chain conversation flow data monitored by a data monitoring port for a supply chain settlement service event, perform abnormal behavior pattern analysis on the first supply chain conversation flow data, and generate a first abnormal behavior pattern analysis result;

[0009] If the first abnormal behavior pattern analysis result reflects that the data monitoring port has abnormal behavior pattern data, executing a data window shielding process for the data monitoring port;

[0010] After executing the data window shielding process, obtaining the second supply chain conversation flow data monitored by the data monitoring port, performing abnormal behavior pattern analysis on the second supply chain conversation flow data, and generating a second abnormal behavior pattern analysis result;

[0011] If the second abnormal behavior pattern analysis result reflects that the abnormal behavior pattern data at the data monitoring port has completed data window shielding, user demand prediction based on the key supply chain conversation flow data monitored from the data monitoring port is performed.

[0012] In a possible implementation manner of the first aspect, performing abnormal behavior pattern analysis on the first supply chain conversation flow data to generate a first abnormal behavior pattern analysis result includes:

[0013] Performing feature extraction on the first supply chain conversation flow data to generate a supply chain conversation path vector graph of the first supply chain conversation flow data;

[0014] Acquire interaction trend characteristic parameters between the data monitoring port and the supply chain conversation element included in the first supply chain conversation flow data;

[0015] Perform abnormal behavior pattern analysis based on the supply chain conversation path vector diagram and the interaction trend characteristic parameters to generate a first abnormal behavior pattern analysis result;

[0016] The step of obtaining the interaction trend characteristic parameters between the data monitoring port and the supply chain conversation elements included in the first supply chain conversation flow data includes:

[0017] Analyzing an activity pattern of each supply chain conversation element in the first supply chain conversation flow data, and extracting time series data related to the supply chain conversation element based on the activity pattern of each supply chain conversation element;

[0018] identifying respective key interaction pattern data from the time series data associated with the supply chain conversation elements;

[0019] Based on the key interaction mode data, the interaction trend characteristic parameters between each supply chain conversation element and the data monitoring port are calculated, and the interaction trend characteristic parameters include at least one of interaction frequency, interaction duration, and interaction stability.

[0020] In a possible implementation manner of the first aspect, the first supply chain conversation flow data includes supply chain order conversation flow data and supply chain product conversation flow data monitored for the same supply chain conversation element; and extracting features from the first supply chain conversation flow data to generate a supply chain conversation path vector diagram of the first supply chain conversation flow data includes:

[0021] Performing feature extraction on the supply chain order conversation flow data and the supply chain product conversation flow data respectively, and generating a supply chain order conversation path vector graph of the supply chain order conversation flow data and a supply chain product conversation path vector graph of the supply chain product conversation flow data;

[0022] The abnormal behavior pattern analysis is performed according to the supply chain conversation path vector diagram and the interaction trend characteristic parameter to generate a first abnormal behavior pattern analysis result, including:

[0023] Fusion of the supply chain order conversation path vector graph, the supply chain product conversation path vector graph, and the interactive trend vector of the interactive trend feature parameter to generate target fusion vector data;

[0024] Performing abnormal behavior pattern analysis based on the target fusion vector data to generate a first abnormal behavior pattern analysis result;

[0025] In a possible implementation of the first aspect, the step of fusing the supply chain order session path vector diagram, the supply chain product session path vector diagram, and the interaction trend vector of the interaction trend feature parameter to generate target fused vector data includes:

[0026] Extracting a first key feature from the supply chain order session path vector diagram, wherein the first key feature includes a path feature, a length feature, and a direction feature of the supply chain order;

[0027] Extracting a second key feature from the supply chain product conversation path vector graph, wherein the second key feature includes a flow path and a transaction volume change of the supply chain product;

[0028] Representing the interaction trend vector of the interaction trend characteristic parameter as a numerical vector sequence, wherein the numerical vector sequence reflects the interaction intensity and trend between the data monitoring port and the supply chain conversation element;

[0029] The first key feature, the second key feature and the numerical vector sequence are fused to generate target fused vector data.

[0030] In a possible implementation of the first aspect, the performing abnormal behavior pattern analysis on the first supply chain conversation flow data to generate a first abnormal behavior pattern analysis result is executed according to an abnormal behavior pattern prediction network; the performing abnormal behavior pattern analysis on the second supply chain conversation flow data to generate a second abnormal behavior pattern analysis result is executed according to the abnormal behavior pattern prediction network; the training step of the abnormal behavior pattern prediction network includes:

[0031] Acquire sample supply chain conversation flow data and sample interaction trend characteristic parameters between the data monitoring port and the supply chain conversation elements included in the sample supply chain conversation flow data; the sample supply chain conversation flow data and the sample interaction trend characteristic parameters have abnormal behavior pattern annotation data;

[0032] Extracting features of the example supply chain conversation flow data according to the initialized abnormal behavior pattern prediction network to generate an example supply chain conversation path vector graph of the example supply chain conversation flow data;

[0033] According to the initialized abnormal behavior pattern prediction network, abnormal behavior pattern analysis is performed according to the example supply chain session path vector diagram and the example interaction trend characteristic parameters to generate an example abnormal behavior pattern analysis result;

[0034] According to the example abnormal behavior pattern analysis result and the abnormal behavior pattern annotation data, the network parameters of the initialized abnormal behavior pattern prediction network are optimized until the network convergence condition is met, and the abnormal behavior pattern prediction network that meets the network convergence condition is generated;

[0035] In a possible implementation manner of the first aspect, the example supply chain conversation flow data includes example supply chain order conversation flow data and example supply chain product conversation flow data; and the extracting features of the example supply chain conversation flow data based on the initialized abnormal behavior pattern prediction network to generate an example supply chain conversation path vector diagram of the example supply chain conversation flow data includes:

[0036] According to the initialized abnormal behavior pattern prediction network, feature extraction is performed on the example supply chain order conversation flow data and the example supply chain product conversation flow data respectively to generate an example supply chain order conversation path vector diagram of the example supply chain order conversation flow data and an example supply chain product conversation path vector diagram of the example supply chain product conversation flow data;

[0037] The abnormal behavior pattern prediction network initialized according to the example supply chain session path vector diagram and the example interaction trend feature parameters is used to perform abnormal behavior pattern analysis to generate an example abnormal behavior pattern analysis result, including:

[0038] Performing feature mapping on the example interaction trend feature parameters according to the initialized abnormal behavior pattern prediction network to generate an example interaction trend vector;

[0039] Merging the example supply chain order conversation path vector graph, the example supply chain product conversation path vector graph, and the example interaction trend vector to generate example target vector data;

[0040] An abnormal behavior pattern analysis is performed based on the example target vector data to generate an example abnormal behavior pattern analysis result.

[0041] In a possible implementation of the first aspect, the method further includes:

[0042] Obtaining a priori user demand forecast data for the supply chain settlement service event;

[0043] Determining predicted feedback data based on the priori user demand prediction data;

[0044] If the first abnormal behavior pattern analysis result reflects that the data monitoring port has abnormal behavior pattern data, executing a data window shielding process for the data monitoring port includes:

[0045] If the first abnormal behavior pattern analysis result reflects that abnormal behavior pattern data exists in the data monitoring port, and the predicted feedback data meets the shielding execution requirement, a data window shielding process for the data monitoring port is executed.

[0046] In a possible implementation of the first aspect, if the first abnormal behavior pattern analysis result reflects that the data monitoring port has abnormal behavior pattern data, executing a data window shielding process for the data monitoring port includes:

[0047] If the first abnormal behavior pattern analysis result reflects that abnormal behavior pattern data exists in the data monitoring port, a data window shielding process is performed for the data monitoring port according to the data window shielding component of the supply chain settlement service event.

[0048] In a possible implementation of the first aspect, the abnormal behavior pattern data at the data monitoring port includes illegal access attempt data; the data window shielding component according to the supply chain settlement service event performs a data window shielding process for the data monitoring port, including:

[0049] The security protection strategy of the supply chain settlement service system is activated to perform data window shielding processing on the illegal access attempt at the data monitoring port through the security protection strategy.

[0050] In a possible implementation of the first aspect, the method further includes:

[0051] If the second abnormal behavior pattern analysis result reflects that there is no abnormal behavior pattern data, determining that the abnormal behavior pattern data at the data monitoring port has completed data window shielding;

[0052] Determine the active range of abnormal data based on the second abnormal behavior pattern analysis result. If the active range of abnormal data is smaller than the active range of abnormal data determined based on the first abnormal behavior pattern analysis result, determine that the abnormal behavior pattern data at the data monitoring port has completed data window shielding.

[0053] In a possible implementation manner of the first aspect, performing user demand forecasting based on the key supply chain conversation flow data monitored from the data monitoring port includes:

[0054] Execute instructions according to user demand prediction to obtain key supply chain conversation flow data monitored from the data monitoring port;

[0055] Matching the key supply chain conversation flow data with predefined historical user demand data to predict user demand based on the key supply chain conversation flow data;

[0056] The historical user demand data includes historical product demand trends and historical product preference characteristics obtained based on analysis of past supply chain conversation flow data;

[0057] The performing data matching analysis on the key supply chain conversation flow data and the pre-stored historical user demand data to predict user demand based on the key supply chain conversation flow data includes:

[0058] Extracting current product demand information and current user preference trend data from the key supply chain conversation flow data;

[0059] Compare and analyze the current product demand information with historical product demand trends to generate product demand forecast results;

[0060] Comparing the current user preference trend data with historical product preference characteristics to generate a user preference prediction result;

[0061] The product demand prediction result and the user preference prediction result are combined to generate a final user demand prediction result.

[0062] On the other hand, an embodiment of the present application also provides a user demand forecasting system based on a supply chain settlement service system, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0063] Based on the above aspects, by analyzing the abnormal behavior pattern of the first supply chain conversation flow data monitored by the data monitoring port, the abnormal behavior pattern data of the data monitoring port can be detected in a timely and accurate manner. Secondly, the data window shielding process is executed to effectively shield the abnormal behavior pattern data to prevent it from interfering with the normal user demand forecast, further ensuring the reliability of the forecast results. Thirdly, by analyzing the abnormal behavior pattern of the second supply chain conversation flow data, the execution effect of the data window shielding process can be verified to ensure that the abnormal behavior data has been completely shielded, providing a clean and accurate data basis for subsequent user demand forecasts. Finally, user demand forecasting based on key supply chain conversation flow data can more accurately predict market dynamics and user needs, provide strong decision support for the supply chain settlement service system, and improve the system's service quality and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a schematic diagram of the execution flow of the user demand prediction method based on the supply chain settlement service system provided in an embodiment of the present application.

[0065] Figure 2 It is a schematic diagram of the hardware architecture of a user demand forecasting system based on a supply chain settlement service system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0066] The present application will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of a user demand prediction method based on a supply chain settlement service system provided by an embodiment of the present application. The user demand prediction method based on a supply chain settlement service system is introduced in detail below.

[0067] Step S110, obtaining first supply chain conversation flow data monitored by a data monitoring port for supply chain settlement service events, performing abnormal behavior pattern analysis on the first supply chain conversation flow data, and generating a first abnormal behavior pattern analysis result.

[0068] In detail, the supply chain settlement service event refers to a specific activity or transaction involving settlement, payment or account processing in the supply chain. These supply chain settlement service events may include order payment, invoice processing, account verification, etc.

[0069] The data monitoring port is an interface for monitoring and collecting supply chain conversation flow data, which can capture and record all interaction data in the supply chain for subsequent analysis. For example, in a supply chain system, the data monitoring port can be a software module that continuously collects data from various supply chain links (such as procurement, sales, logistics, etc.), including order information, product flow records, etc.

[0070] The first supply chain conversation flow data refers to the supply chain interaction data captured for the first time through the data monitoring port. These first supply chain conversation flow data are real-time and reflect the current activity status in the supply chain. For example, in the supply chain system of an e-commerce platform, the first supply chain conversation flow data may include data on a series of operations such as user browsing products, adding to shopping carts, placing orders and paying.

[0071] Abnormal behavior pattern analysis is a data analysis method used to identify abnormal behaviors in supply chain conversation flow data that do not conform to normal patterns. These anomalies may indicate errors, fraud, or other non-normal activities.

[0072] In this embodiment, the server receives an instruction to obtain data from the data monitoring port of the supply chain settlement service event. The server immediately establishes a connection with the data monitoring port and starts to receive the first supply chain session flow data, which includes detailed information on various supply chain transaction activities, such as order creation, product flow, payment settlement, etc.

[0073] After receiving the first supply chain conversation flow data, the server performs abnormal behavior pattern analysis. First, the server extracts features from the first supply chain conversation flow data and generates a supply chain conversation path vector map, which can clearly show the path and nodes of data flow. Then, the server analyzes the interaction trend between the first supply chain conversation flow data and the data monitoring port, and calculates parameters such as interaction frequency and interaction duration.

[0074] Through these analyses, the server discovered an abnormal pattern: in a certain period of time, the frequency of interaction between the data monitoring port and a supply chain conversation element was abnormally high and lasted for a long time. The server immediately recorded this abnormal pattern and generated the first abnormal behavior pattern analysis result.

[0075] Step S120: If the first abnormal behavior pattern analysis result reflects that the data monitoring port has abnormal behavior pattern data, a data window shielding process for the data monitoring port is executed.

[0076] In detail, when an abnormal behavior pattern is detected, in order to protect the system from potential threats, a process is initiated to mask or isolate these abnormal data, which is called the data window masking process.

[0077] In this embodiment, according to the first abnormal behavior pattern analysis result, the server determines that there is abnormal behavior pattern data in the data monitoring port. In order to ensure the security and stability of the supply chain settlement service, the server decides to immediately execute the data window shielding process.

[0078] The server first determines the source and scope of the abnormal data, and then activates the security protection strategy, which can establish a temporary data shielding window at the data monitoring port to prevent further inflow of abnormal data. At the same time, the server can also send an alert to the administrator to notify the occurrence of the abnormal situation and the progress of the processing.

[0079] Step S130, after executing the data window shielding process, obtaining the second supply chain conversation flow data monitored by the data monitoring port, performing abnormal behavior pattern analysis on the second supply chain conversation flow data, and generating a second abnormal behavior pattern analysis result.

[0080] After executing the data window shielding process, continue to monitor and collect new supply chain session data, which are called second supply chain session flow data. These second supply chain session flow data should not contain the abnormal behavior patterns detected previously. That is, after shielding the abnormal order requests, other normal order data that continues to be collected belongs to the second supply chain session flow data.

[0081] In this embodiment, after executing the data window shielding process, the server obtains the supply chain session flow data from the data monitoring port again. This time, the second supply chain session flow data is obtained. These second supply chain session flow data are monitored after the shielding process is executed, and therefore should not contain the previous abnormal behavior pattern data.

[0082] To verify this, the server performed the same abnormal behavior pattern analysis on the second supply chain conversation flow data. The server again generated the supply chain conversation path vector diagram and interaction trend feature parameters, and carefully analyzed these second supply chain conversation flow data.

[0083] Step S140: If the second abnormal behavior pattern analysis result reflects that the abnormal behavior pattern data at the data monitoring port has completed data window shielding, user demand prediction is performed based on the key supply chain conversation flow data monitored from the data monitoring port.

[0084] Specifically, the key supply chain conversation flow data refers to the part of the supply chain conversation flow data that is crucial for user demand prediction, which may include the user's purchase history, browsing habits, etc., and is crucial for predicting the user's future needs and behaviors. For example, on an e-commerce platform, the data of a certain type of goods that users frequently browse and purchase belong to the key supply chain conversation flow data, because this data can help predict the user's future demand for this type of goods.

[0085] That is, in this embodiment, after analysis, the server finds that there is no previous abnormal behavior pattern data in the second supply chain conversation flow data, which means that the data window shielding process has successfully prevented the inflow of abnormal data. Since the data monitoring port has returned to normal, the server can continue to perform normal user demand forecasting. The server extracts key supply chain conversation flow data from the second supply chain conversation flow data, and matches and analyzes these key supply chain conversation flow data with historical user demand data. Through analysis, the server predicts product demand and user preference trends in the future. For example, based on the product flow path and transaction volume changes in the supply chain, the server predicts that a certain product will be favored by users in the near future, and these user demand forecast results will be used to guide the management and optimization of the supply chain.

[0086] Based on the above steps, by analyzing the abnormal behavior pattern of the first supply chain conversation flow data monitored by the data monitoring port, the abnormal behavior pattern data of the data monitoring port can be detected in a timely and accurate manner. Secondly, the data window shielding process is executed, which can effectively shield the abnormal behavior pattern data to prevent it from interfering with the normal user demand forecast, further ensuring the reliability of the forecast results. Thirdly, by analyzing the abnormal behavior pattern of the second supply chain conversation flow data, the execution effect of the data window shielding process can be verified to ensure that the abnormal behavior data has been completely shielded, providing a clean and accurate data basis for subsequent user demand forecasts. Finally, user demand forecasting based on key supply chain conversation flow data can more accurately predict market dynamics and user needs, provide strong decision support for the supply chain settlement service system, and improve the system's service quality and user satisfaction.

[0087] In a possible implementation, step S110 may include:

[0088] Step S111 : extracting features from the first supply chain conversation flow data to generate a supply chain conversation path vector diagram of the first supply chain conversation flow data.

[0089] Step S112: Acquire interaction trend characteristic parameters between the data monitoring port and the supply chain conversation elements included in the first supply chain conversation flow data.

[0090] Step S113, performing abnormal behavior pattern analysis based on the supply chain conversation path vector diagram and the interaction trend characteristic parameters to generate a first abnormal behavior pattern analysis result.

[0091] In this embodiment, the server receives first supply chain conversation flow data from various supply chain links (such as procurement, production, logistics, etc.). These first supply chain conversation flow data record in detail the information exchange between each link, including key information such as timestamp, sender, receiver, message content, etc.

[0092] The server will first pre-process the first supply chain conversation flow data, clean and organize the data format to ensure the accuracy of subsequent analysis. Next, the server begins to extract features from the first supply chain conversation flow data. During the feature extraction process, the server can focus on key features such as the start point, end point, frequency, duration, and transition probability of the conversation.

[0093] After completing the feature extraction, the server can generate a supply chain conversation path vector diagram based on these features. The supply chain conversation path vector diagram can clearly display the interaction path and interaction intensity of each link in the supply chain, providing an intuitive and effective visualization tool for the subsequent identification of abnormal behavior patterns.

[0094] After generating the supply chain session path vector diagram, the server can further analyze the interaction between the data monitoring port and the supply chain session elements. The data monitoring port here can be understood as the interface used by the server to collect and monitor supply chain data, while the supply chain session elements refer to the various components in the session flow data, such as sender, receiver, message type, etc.

[0095] The server can calculate and extract the interaction trend characteristic parameters between the data monitoring port and each supply chain session element. These interaction trend characteristic parameters may include but are not limited to: interaction frequency, interaction duration, data transmission volume, etc. These interaction trend characteristic parameters can reflect the activity level and interaction mode between various links in the supply chain, and provide important data support for subsequent abnormal behavior pattern analysis.

[0096] After obtaining the supply chain conversation path vector diagram and interaction trend characteristic parameters, the server can use these data to analyze abnormal behavior patterns, thereby generating a first abnormal behavior pattern analysis result, which can list in detail all behavior patterns identified as abnormal.

[0097] Wherein, step S112 may include:

[0098] Step S1121 , analyzing the activity pattern of each supply chain conversation element in the first supply chain conversation flow data, and extracting time series data related to the supply chain conversation element based on the activity pattern of each supply chain conversation element.

[0099] Step S1122, identifying each key interaction pattern data from the time series data related to the supply chain conversation element.

[0100] Step S1123, based on the key interaction mode data, calculate the interaction trend characteristic parameters between each supply chain conversation element and the data monitoring port, and the interaction trend characteristic parameters include at least one of interaction frequency, interaction duration, and interaction stability.

[0101] In this embodiment, the server starts to process the first supply chain conversation flow data, which records the information exchange between the participants in the supply chain (such as suppliers, manufacturers, distributors, etc.). The server first analyzes the activity pattern of the supply chain conversation elements in the first supply chain conversation flow data.

[0102] Specifically, the server can pay attention to when each session element starts to be active, the duration of the activity, the frequency of the activity, etc. For example, a supplier may be active between 9 am and 5 pm on weekdays, while a distributor may be more active in the afternoon and evening because they are processing orders and shipments during the day.

[0103] By analyzing these activity patterns, the server can build a behavioral model of each supply chain conversation element, laying the foundation for subsequent data extraction and pattern recognition.

[0104] After analyzing the activity patterns of each supply chain session element, the server begins to extract relevant time series data from the first supply chain session flow data according to these patterns. Time series data is a series of data points arranged in chronological order. For supply chain session elements, these data points may include the start time, end time, session duration, etc. of each session. The server may arrange these data points in chronological order to form a complete time series.

[0105] For example, for a specific supplier, the server can extract the start and end time of all sessions with the supplier, as well as the duration of each session, to form a time series data about the supplier.

[0106] After extracting the time series data related to the supply chain conversation elements, the server can begin to identify the key interaction pattern data. Key interaction patterns refer to the interaction methods that appear frequently in the supply chain conversation and have specific meanings or impacts. For example, there may be a fixed interaction pattern between a supplier and a distributor: at 3 pm every day, the supplier will send an inventory update report to the distributor, and the distributor will confirm and reply within 1 hour after receiving the report. The server can automatically identify and extract these key interaction pattern data through algorithms to provide a basis for the subsequent calculation of interaction trend feature parameters.

[0107] After identifying the key interaction pattern data, the server can start calculating the interaction trend characteristic parameters between each supply chain conversation element and the data monitoring port. These interaction trend characteristic parameters include interaction frequency, interaction duration, and interaction stability. Interaction frequency refers to the number of interactions per unit time; interaction duration refers to the average length of each interaction; and interaction stability reflects the stability and consistency of the interaction pattern. For example, for the fixed interaction pattern between suppliers and distributors mentioned earlier, the server can calculate their interaction frequency (such as once a day), interaction duration (such as an average of 10 minutes per interaction), and interaction stability (such as uninterrupted interaction for 30 consecutive days). These parameters help the server comprehensively evaluate the interaction between supply chain conversation elements and data monitoring ports, and provide an important reference for subsequent abnormal behavior detection.

[0108] In a possible implementation, the first supply chain conversation flow data includes supply chain order conversation flow data and supply chain product conversation flow data monitored for the same supply chain conversation element. Step S111 may include: performing feature extraction on the supply chain order conversation flow data and the supply chain product conversation flow data, respectively, to generate a supply chain order conversation path vector diagram of the supply chain order conversation flow data and a supply chain product conversation path vector diagram of the supply chain product conversation flow data.

[0109] In this embodiment, the server receives two groups of first supply chain conversation flow data: one group is conversation data for supply chain order communication (supply chain order conversation flow data), and the other group is conversation data for supply chain product information communication (supply chain product conversation flow data).

[0110] For supply chain order conversation flow data, the server can analyze the key nodes in the entire process from order creation to completion, such as order submission, confirmation, shipment, receipt, etc., and record the flow relationship and timestamps between these nodes. Through feature extraction, the server can identify the main paths and common patterns in the order processing process, and then generate a supply chain order conversation path vector diagram, which clearly shows the flow path of order data between different links.

[0111] For supply chain product conversation flow data, the server can focus on product-related information exchanges, such as product specifications, quality, inventory status, etc. Through feature extraction, the server can identify the main participants and communication patterns in product information exchanges, and generate a supply chain product conversation path vector map based on this, which depicts the transmission path of product information between various links in the supply chain.

[0112] After completing the feature extraction, the server can generate two vector graphs based on the extracted features: a supply chain order session path vector graph and a supply chain product session path vector graph.

[0113] The supply chain order session path vector diagram graphically shows the flow path and key nodes of order data in the supply chain. For example, the diagram may include the order data transmission path from supplier to manufacturer and then to distributor, as well as the processing time and status changes at each node.

[0114] The supply chain product conversation path vector diagram shows the transmission path of product information in the supply chain. This diagram can help the server understand the flow pattern of product information, such as which links frequently communicate product information and how these exchanges affect order processing.

[0115] Step S112 may include: fusing the supply chain order conversation path vector diagram, the supply chain product conversation path vector diagram and the interaction trend vector of the interaction trend feature parameter to generate target fusion vector data, performing abnormal behavior pattern analysis based on the target fusion vector data to generate a first abnormal behavior pattern analysis result.

[0116] In this embodiment, after generating the supply chain order session path vector diagram and the supply chain product session path vector diagram, the server will fuse these vector data with the interaction trend vector of the interaction trend feature parameter calculated previously. The interaction trend feature parameter reflects the interaction trend between the data monitoring port and the supply chain session element, including the interaction frequency, interaction duration, etc. These parameters can be converted into the interaction trend vector to facilitate mathematical operations and fusion with other vector data.

[0117] The fusion process may involve the allocation of weights and the selection of algorithms to ensure that data from different sources can be reasonably reflected in the target fused vector data. The server may select a suitable fusion algorithm to fuse the three vector data (supply chain order conversation path vector diagram, supply chain product conversation path vector diagram, and interaction trend vector) into one target fused vector data.

[0118] After obtaining the target fusion vector data, the server can use this data to analyze abnormal behavior patterns. By comparing normal behavior patterns (these patterns can be learned from historical data), the server can identify behaviors that deviate greatly from normal patterns, i.e. abnormal behaviors.

[0119] For example, if the target fusion vector data shows that the interaction frequency of a certain link has suddenly increased significantly, or the flow path of the order data has unusual branches, these may be regarded as abnormal behavior patterns. The server can further analyze the reasons behind these abnormal patterns, such as possible fraud, system failures, or other risk factors in the supply chain.

[0120] Finally, the server can generate a first abnormal behavior pattern analysis result, which will list in detail all behavior patterns identified as abnormal, as well as possible causes and potential impacts. In this way, relevant personnel can take corresponding measures in a timely manner based on this result to ensure the security and stability of the supply chain.

[0121] In a possible implementation, the step of fusing the supply chain order conversation path vector diagram, the supply chain product conversation path vector diagram, and the interactive trend vector of the interactive trend feature parameter to generate target fused vector data includes:

[0122] Step S1121, extracting a first key feature from the supply chain order session path vector map, wherein the first key feature includes a path feature, a length feature, and a direction feature of the supply chain order.

[0123] Step S1122: extracting a second key feature from the supply chain product session path vector map, wherein the second key feature includes a flow path and a transaction volume change of the supply chain product.

[0124] Step S1123, representing the interaction trend vector of the interaction trend characteristic parameter as a numerical vector sequence, wherein the numerical vector sequence reflects the interaction intensity and trend between the data monitoring port and the supply chain conversation element.

[0125] Step S1124, fusing the first key feature, the second key feature and the numerical vector sequence to generate target fused vector data.

[0126] In this embodiment, the server first processes the supply chain order session path vector map, which is generated by the previous steps and shows in detail the flow path of the order data in the supply chain. The server begins to extract the first key features from this supply chain order session path vector map, which include the path features of the supply chain order (such as the starting point, end point and nodes passed by the order), length features (the total length of the order path or the distance between each node) and direction features (the direction of the order data flow).

[0127] For example, the server may recognize that a typical order path is from a supplier to a manufacturer and then to a distributor. The length feature of the path can be the sum of the distances between the three nodes, while the direction feature indicates that the data flows from the supplier to the distributor.

[0128] Next, the server processes the supply chain product session path vector map, which depicts the delivery path of product information in the supply chain.

[0129] The server extracts the second key feature from this supply chain product session path vector diagram, which mainly includes the flow path of the supply chain product (the entire process from product production to sales) and the change in transaction volume (change in the number or frequency of product transactions in each link).

[0130] For example, the server may find that the transaction volume of a product suddenly increases at a certain point in time during its flow path from the manufacturer to the distributor, which may indicate that the product is popular in the market or there is a promotion activity.

[0131] The server has calculated the interaction trend characteristic parameters between the data monitoring port and the supply chain conversation elements, such as interaction frequency, interaction duration, etc.

[0132] Now, the server converts these interaction trend characteristic parameters into a numerical vector sequence, which reflects the interaction intensity and trend between the data monitoring port and the supply chain conversation elements in the form of numerical values. For example, a high-frequency interaction can be represented as a higher numerical value, while a low-frequency interaction corresponds to a lower numerical value.

[0133] Finally, the server fuses the first key features (order path, length and direction features), the second key features (product flow path and transaction volume changes) and the numerical vector sequence (numerical representation of interactive trends) extracted earlier. The fusion process can involve complex algorithms and data processing techniques, such as weighted average and principal component analysis, to ensure that data from different sources can be reasonably combined.

[0134] The fused data is called target fusion vector data, which integrates multiple aspects of supply chain orders, products and interaction trends, providing a rich information basis for subsequent analysis of abnormal behavior patterns.

[0135] In a possible implementation, the abnormal behavior pattern analysis of the first supply chain conversation flow data to generate a first abnormal behavior pattern analysis result is executed according to an abnormal behavior pattern prediction network. The abnormal behavior pattern analysis of the second supply chain conversation flow data to generate a second abnormal behavior pattern analysis result is executed according to the abnormal behavior pattern prediction network. The training step of the abnormal behavior pattern prediction network includes:

[0136] Step S101, obtaining sample supply chain conversation flow data and sample interaction trend characteristic parameters between the data monitoring port and the supply chain conversation elements included in the sample supply chain conversation flow data. The sample supply chain conversation flow data and the sample interaction trend characteristic parameters have abnormal behavior pattern annotation data.

[0137] Step S102 , extracting features from the example supply chain conversation flow data according to the initialized abnormal behavior pattern prediction network, and generating an example supply chain conversation path vector diagram of the example supply chain conversation flow data.

[0138] Step S103, performing abnormal behavior pattern analysis based on the initialized abnormal behavior pattern prediction network and the example supply chain session path vector diagram and the example interaction trend feature parameters to generate an example abnormal behavior pattern analysis result.

[0139] Step S104, optimizing network parameters of the initialized abnormal behavior pattern prediction network according to the abnormal behavior pattern analysis result of the example and the abnormal behavior pattern annotation data until a network convergence condition is met, thereby generating the abnormal behavior pattern prediction network that meets the network convergence condition.

[0140] In this embodiment, the server obtains a set of sample supply chain conversation flow data from the database. These data record various activities in previous supply chain conversations, including order processing, product information exchange, etc. At the same time, the server also obtains sample interaction trend characteristic parameters related to these data, such as interaction frequency, interaction duration, etc. Importantly, these data have been marked to see if there are abnormal behavior patterns, which will provide an important reference for subsequent network training.

[0141] The server uses the initialized abnormal behavior pattern prediction network to extract features from the sample supply chain conversation flow data. This process mainly identifies key information in the conversation flow data, such as the starting point, end point, key interaction nodes, etc. of the conversation, and represents this information in a graphical way to generate a sample supply chain conversation path vector diagram, which can clearly show the process and path of the supply chain conversation.

[0142] The server inputs the sample supply chain conversation path vector diagram and the sample interaction trend characteristic parameters into the initialized abnormal behavior pattern prediction network to analyze the abnormal behavior pattern. The network will judge whether there is an abnormal behavior pattern based on these input data, combined with its internal algorithms and models, and generate a sample abnormal behavior pattern analysis result, which will list in detail all the behavior patterns identified as abnormal and their related features.

[0143] The server compares the analysis results of the example abnormal behavior pattern with the known abnormal behavior pattern annotation data and calculates the difference between the two. Then, based on this difference, the parameters of the initialized abnormal behavior pattern prediction network are optimized, and the network's weights, biases and other parameters are adjusted to improve the network's ability to recognize abnormal behavior patterns. This process is repeated until the network meets the convergence conditions, that is, the network's performance reaches the optimal level or no longer improves significantly. Finally, the server can generate an abnormal behavior pattern prediction network that meets the network convergence conditions, which will be used in subsequent supply chain conversation flow data abnormal behavior pattern analysis tasks.

[0144] In a possible implementation, the example supply chain conversation flow data includes example supply chain order conversation flow data and example supply chain product conversation flow data. Step S102 may include:

[0145] Step S1021, based on the initialized abnormal behavior pattern prediction network, feature extraction is performed on the example supply chain order conversation flow data and the example supply chain product conversation flow data respectively, to generate an example supply chain order conversation path vector diagram of the example supply chain order conversation flow data and an example supply chain product conversation path vector diagram of the example supply chain product conversation flow data.

[0146] Step S103 may include:

[0147] Step S1031 , performing feature mapping on the example interaction trend feature parameters according to the initialized abnormal behavior pattern prediction network to generate a example interaction trend vector.

[0148] Step S1032 , fusing the example supply chain order conversation path vector map, the example supply chain product conversation path vector map, and the example interaction trend vector map to generate example target vector data.

[0149] Step S1033 , performing abnormal behavior pattern analysis based on the example target vector data to generate an example abnormal behavior pattern analysis result.

[0150] In this embodiment, the server first processes the sample supply chain order conversation flow data, which may contain the complete process records of historical orders from creation, confirmation, delivery to receipt. The server uses the initialized abnormal behavior pattern prediction network to extract features from these sample supply chain order conversation flow data and identify key nodes and flow relationships in the order processing process, such as order submission time, confirmation time, delivery time, etc. Through these features, the server generates a sample supply chain order conversation path vector diagram, which graphically shows the flow path of order data in the supply chain.

[0151] Similarly, the server also extracts features from the sample supply chain product conversation flow data, which may include product specifications, inventory status, transaction records, etc. The server analyzes this information to identify the main participants and communication patterns of product information exchange, and generates a sample supply chain product conversation path vector diagram based on this information.

[0152] The server obtains the characteristic parameters of the example interaction trend related to the example supply chain conversation flow data, such as interaction frequency, interaction duration, etc. These parameters reflect the interaction trend between the data monitoring port and the supply chain conversation elements. In order to effectively integrate these parameters with the supply chain conversation path vector map, the server uses the initialized abnormal behavior pattern prediction network to perform feature mapping on these parameters and convert them into example interaction trend vectors, which represent the intensity and direction of the interaction trend in numerical form.

[0153] After obtaining the sample supply chain order conversation path vector diagram, the sample supply chain product conversation path vector diagram, and the sample interaction trend vector diagram, the server begins to fuse these data. The fusion process may involve the allocation of weights and the selection of algorithms to ensure that data from different sources can be reasonably reflected in the fused data. Through the fusion algorithm, the server integrates these vector data into a sample target vector data.

[0154] Finally, the server uses the initialized abnormal behavior pattern prediction network to perform abnormal behavior pattern analysis on the example target vector data. By comparing normal behavior patterns (these patterns can be learned from historical data or defined by expert knowledge), the server can identify behaviors that deviate greatly from normal patterns, namely abnormal behavior patterns. For example, if the example target vector data shows that the interaction frequency of a certain link suddenly increases significantly or the flow path of the order data has unusual branches, these may be regarded as abnormal behavior patterns. The server can further analyze the reasons behind these abnormal patterns and generate example abnormal behavior pattern analysis results, which will list in detail all behavior patterns identified as abnormal, as well as possible causes and potential impacts, providing important reference for subsequent optimization and adjustment.

[0155] In a possible implementation, the method further includes:

[0156] Step A110, obtaining a priori user demand forecast data for the supply chain settlement service event.

[0157] Step A120, determining predicted feedback data based on the priori user demand prediction data.

[0158] Step S120 may include: if the first abnormal behavior pattern analysis result reflects that the data monitoring port has abnormal behavior pattern data, and the predicted feedback data meets the shielding execution requirement, executing a data window shielding process for the data monitoring port.

[0159] In this embodiment, the server first connects to a database containing user demand forecast data, which stores user demand data related to historical supply chain settlement service events and forecast models generated based on these data. The server obtains the prior user demand forecast data about supply chain settlement service events in the future from this database through a specific query statement. These data may include information such as the expected transaction volume, user activity forecast, and user preference for settlement services.

[0160] After the server obtains the priori user demand forecast data, it starts to analyze the priori user demand forecast data. First, the data is cleaned and preprocessed to remove outliers and noise to ensure the accuracy and reliability of the data.

[0161] Next, the server conducts in-depth analysis and calculations on these pre-processed prior user demand forecast data. During this process, the server can take into account a variety of factors, such as trends in historical data, seasonal changes, changes in the market environment, etc. Finally, based on these analyses and calculations, the server generates forecast feedback data, which are the forecast results of user demand for supply chain settlement service events in the future, and can include the expected growth trend of transaction volume, the trend of changes in user activity, etc.

[0162] Therefore, the server has generated a first abnormal behavior pattern analysis result through the previous steps, and this result reflects whether there is abnormal behavior pattern data on the data monitoring port.

[0163] Now, the server starts to determine whether to execute the data window shielding process based on this result. First, check the first abnormal behavior pattern analysis result. If abnormal behavior pattern data is found in the data monitoring port, then further check the predicted feedback data.

[0164] The server can determine whether the forecast feedback data meets the shielding execution requirements, which can be a preset threshold or a series of complex conditional judgment logic. If the forecast feedback data meets these requirements, such as the predicted transaction volume growth trend exceeds a certain threshold, or the change trend of user activity is consistent with the abnormal behavior pattern data, then the server will decide to execute the data window shielding process.

[0165] When executing the data window shielding process, the server can close or restrict access to the data monitoring port to prevent abnormal behavior pattern data from adversely affecting supply chain settlement service events. At the same time, the server may also trigger an alarm system to notify relevant personnel to handle these abnormal data in a timely manner.

[0166] In a possible implementation, step S120 may further include:

[0167] If the first abnormal behavior pattern analysis result reflects that abnormal behavior pattern data exists in the data monitoring port, a data window shielding process is performed for the data monitoring port according to the data window shielding component of the supply chain settlement service event.

[0168] In this embodiment, after detecting that the first abnormal behavior pattern analysis result reflects that there is abnormal behavior pattern data in the data monitoring port, the server will immediately activate the response mechanism. A key link in the response mechanism is to call the data window shielding component of the supply chain settlement service event.

[0169] This data window shielding component is pre-configured to handle such abnormal situations. When the server decides to start the data window shielding process, it will first communicate with this component to pass detailed information about the abnormal behavior pattern data, including the type of abnormality, severity, and possible impact range. Next, the data window shielding component will formulate a shielding strategy based on this information, which may include temporarily shutting down certain functions of the data monitoring port, restricting access to specific data, or conducting stricter review and filtering of data transmitted through the port.

[0170] Once the shielding strategy is determined, the data window shielding component will send instructions to the server, and the server will perform specific data window shielding operations based on these instructions, which may involve modifying the network configuration, updating the access control list, adjusting the data filtering rules, etc.

[0171] While executing the data window shielding process, the server will also record all operation logs for subsequent analysis and auditing. In addition, if the shielding operation has an impact on the supply chain settlement service event, the server will promptly notify the relevant users and system administrators to ensure that they understand the current situation and take appropriate countermeasures.

[0172] Through this data window shielding process, the server can effectively isolate and process abnormal behavior pattern data, thereby protecting the security and stability of supply chain settlement service events.

[0173] In a possible implementation, the abnormal behavior pattern data at the data monitoring port includes illegal access attempt data. The data window shielding component based on the supply chain settlement service event performs a data window shielding process for the data monitoring port, including: activating the security protection strategy of the supply chain settlement service system to perform data window shielding processing on the illegal access attempt at the data monitoring port through the security protection strategy.

[0174] In this embodiment, when the server detects abnormal behavior pattern data, especially illegal access attempt data, on the data monitoring port, it will immediately take action to protect the security of the system. One of the key steps is to activate the security protection strategy of the supply chain settlement service system.

[0175] This security protection strategy is pre-set by the system to deal with various network security threats, including illegal access attempts. When the server identifies illegal access attempt data, a security alarm is triggered and the corresponding security protection strategy is automatically activated.

[0176] After activating the security protection policy, the server can first configure and optimize the policy to ensure that it can effectively respond to current illegal access attempts. This can include updating firewall rules, strengthening user authentication mechanisms, and restricting access to sensitive data.

[0177] Once the security protection policy is configured, the server can start to perform data window shielding, which means that the server can intercept and filter the data stream from illegal access attempts to prevent malicious data from entering the system. At the same time, the server will also clean and isolate the data that has entered the system to ensure the security and stability of the system.

[0178] During the data window shielding process, the server can continuously monitor and analyze the data flow at the data monitoring port to promptly detect and handle any new illegal access attempts. If any suspicious activities are found, the server can immediately take corresponding measures, such as blocking the source IP address, issuing security warnings, etc.

[0179] Through this process, the server can effectively protect the supply chain settlement service system from the threat of illegal access, ensuring the normal operation of the system and the security of data.

[0180] In a possible implementation, the method further includes:

[0181] Step B110: If the second abnormal behavior pattern analysis result reflects that there is no abnormal behavior pattern data, it is determined that the abnormal behavior pattern data at the data monitoring port has completed data window shielding.

[0182] In this embodiment, after the data window shielding process is executed for a period of time, the server can run the abnormal behavior pattern analysis again, and the result of this analysis is called the second abnormal behavior pattern analysis result. If the result of this analysis shows that there is no abnormal behavior pattern data at the data monitoring port, it means that the abnormal behavior pattern data detected previously may have been successfully shielded or processed.

[0183] For example, the server detected a large number of illegal access attempts to the data monitoring port at a certain point in time, thus triggering the data window shielding process. Now, after a period of time, the server performs abnormal behavior detection again and finds that there are no signs of illegal access, which indicates that the data window shielding process may have effectively prevented illegal access, so the server can determine that the abnormal behavior pattern data at the data monitoring port has completed the data window shielding.

[0184] Step B120, determining the active range of abnormal data based on the second abnormal behavior pattern analysis result; if the active range of abnormal data is smaller than the active range of abnormal data determined based on the first abnormal behavior pattern analysis result, determining that the abnormal behavior pattern data at the data monitoring port has completed data window shielding.

[0185] In addition to checking whether there is abnormal behavior pattern data in the second abnormal behavior pattern analysis result, the server also analyzes the active range of the abnormal data in this analysis result. The active range may include the time period when the abnormal data appears, the number of IP addresses involved, the frequency of attempted access, etc.

[0186] If the second abnormal behavior pattern analysis result shows that the active range of abnormal data is significantly smaller than the range in the first analysis result, this usually means that the data window shielding process has effectively limited the spread of abnormal behavior. For example, the first analysis result may show that abnormal data frequently appears in multiple time periods and involves a large number of IP addresses; while the second analysis result may only show a small amount of abnormal data in a very short period of time, and the IP addresses involved are also greatly reduced.

[0187] In this case, the server can reasonably infer that the abnormal behavior pattern data at the data monitoring port has been effectively shielded. Because the active range of the abnormal data has been reduced, it is direct evidence of the success of the data window shielding process. The server can therefore determine that the abnormal behavior pattern data at the data monitoring port has completed the data window shielding, and may further adjust or end the current shielding process.

[0188] In a possible implementation, step S140 may include:

[0189] Step S141, executing instructions according to user demand prediction, obtaining key supply chain conversation flow data monitored from the data monitoring port.

[0190] Step S142 : matching the key supply chain conversation flow data with predefined historical user demand data to perform user demand prediction based on the key supply chain conversation flow data.

[0191] In this embodiment, the server receives a user demand forecast execution instruction, which can be manually triggered by a system administrator or automatically sent according to a preset scheduled task. After receiving the user demand forecast execution instruction, the server starts to obtain key supply chain session flow data from the data monitoring port.

[0192] These key supply chain conversation flow data may include user interaction records with the supply chain system, product query and browsing history, shopping cart content, transaction records, etc. The server captures this data from the data monitoring port in real time or periodically through a specific data interface or network protocol, and stores it in an internal database for subsequent analysis.

[0193] After acquiring the key supply chain conversation flow data, the server begins to match these key supply chain conversation flow data with the pre-stored historical user demand data. The historical user demand data is obtained after in-depth analysis of past supply chain conversation flow data, which includes important information such as historical product demand trends and historical product preference characteristics.

[0194] The matching process is achieved through specific algorithms and models, such as machine learning algorithms that can be used to identify the similarities and correlations between current session flow data and historical data. In this way, the server can more accurately understand the user's current needs and preferences.

[0195] The historical user demand data includes historical product demand trends and historical product preference characteristics obtained based on analysis of past supply chain conversation flow data.

[0196] Step S142 includes:

[0197] Step S1421, extracting current product demand information and current user preference trend data from the key supply chain conversation flow data.

[0198] Step S1422, comparing and analyzing the current product demand information with historical product demand trends to generate product demand forecast results.

[0199] Step S1423, comparing the current user preference trend data with historical product preference features to generate a user preference prediction result.

[0200] Step S1424, combining the product demand prediction result and the user preference prediction result to generate a final user demand prediction result.

[0201] In this embodiment, in detail, during the data matching process, the server can extract current product demand information and user preference trend data from the key supply chain conversation flow data. Product demand information may include specific product types, brands, price ranges, etc. that users query or browse. User preference trend data may reflect the user's preference for certain product features, such as color, size, function, etc.

[0202] This information is obtained by analyzing user behavior patterns in conversation flow data. For example, the time a user stays on a product page, the links they click, the products they add to the shopping cart, etc. can all be used as a basis for refining this information.

[0203] After extracting the current product demand information, the server can compare and analyze this information with the historical product demand trends. For example, the server can check the sales changes of similar products in different time periods, user feedback and other information in the historical data to predict the demand trend of the current product.

[0204] Through comparative analysis, the server can generate a product demand forecast result, which can include key indicators such as the product's expected sales volume and popularity, and has important guiding significance for supply chain management and inventory planning.

[0205] At the same time, the server will also compare the current user preference trend data with the historical product preference characteristics. The comparison process is mainly to identify changes and trends in user preferences in order to better meet the user's personalized needs.

[0206] For example, if a user recently shows a stronger preference for products made of environmentally friendly materials, the server will compare this information with historical data to see if this preference is a new trend or just an accidental behavior of individual users. Through comparison and analysis, the server can generate a user preference prediction result, which helps companies adjust product design and market strategies to better cater to user needs.

[0207] After generating the product demand forecast results and user preference forecast results respectively, the server can combine these two results to generate the final user demand forecast result, which is a comprehensive, multi-dimensional analysis report that not only includes key indicators such as the product's expected sales and popularity, but also reflects the user's personalized needs and preference trends.

[0208] The final user demand forecast result is of great value to enterprises, helping them to more accurately grasp market dynamics and user needs, and thus make more informed decisions and plans. For example, enterprises can adjust production plans, optimize inventory management, improve product design, etc. based on this result to improve market competitiveness and user satisfaction.

[0209] Figure 2 The hardware structure of the user demand forecasting system 100 based on the supply chain settlement service system for implementing the user demand forecasting method based on the supply chain settlement service system provided in the embodiment of the present application is shown as follows: Figure 2As shown, the user demand prediction system 100 based on the supply chain settlement service system may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0210] In one possible design, the user demand forecasting system 100 based on the supply chain settlement service system can be a single server or a server group. The server group can be centralized or distributed (for example, the user demand forecasting system 100 based on the supply chain settlement service system can be a distributed system). In some embodiments, the user demand forecasting system 100 based on the supply chain settlement service system can be local or remote. For example, the user demand forecasting system 100 based on the supply chain settlement service system can access information and / or data stored in the machine-readable storage medium 120 via a network. For another example, the user demand forecasting system 100 based on the supply chain settlement service system can be directly connected to the machine-readable storage medium 120 to access the stored information and / or data. In some embodiments, the user demand forecasting system 100 based on the supply chain settlement service system can be implemented on a user demand forecasting system based on the supply chain settlement service system. As an example only, the user demand forecasting system based on the supply chain settlement service system can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, etc. or any combination thereof.

[0211] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data obtained from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions that the user demand forecasting system 100 based on the supply chain settlement service system uses to execute or use to complete the exemplary method described in this application.

[0212] During the specific implementation process, at least one processor 110 executes computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the user demand prediction method based on the supply chain settlement service system in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0213] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the user demand forecasting system 100 based on the supply chain settlement service system mentioned above. The implementation principles and technical effects are similar and will not be repeated in this embodiment.

[0214] In addition, an embodiment of the present application also provides a readable storage medium, in which computer executable instructions are preset. When the processor executes the computer executable instructions, the user demand prediction method based on the supply chain settlement service system as described above is implemented.

[0215] It should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof. Similarly, it should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof.

Claims

1. A user demand forecasting method based on a supply chain settlement service system, characterized in that: The method comprises: Acquire first supply chain conversation flow data monitored by a data monitoring port for a supply chain settlement service event, perform abnormal behavior pattern analysis on the first supply chain conversation flow data, and generate a first abnormal behavior pattern analysis result; If the first abnormal behavior pattern analysis result reflects that the data monitoring port has abnormal behavior pattern data, executing a data window shielding process for the data monitoring port; After executing the data window shielding process, obtaining the second supply chain conversation flow data monitored by the data monitoring port, performing abnormal behavior pattern analysis on the second supply chain conversation flow data, and generating a second abnormal behavior pattern analysis result; If the second abnormal behavior pattern analysis result reflects that the abnormal behavior pattern data at the data monitoring port has completed data window shielding, based on the key supply chain conversation flow data monitored from the data monitoring port, user demand prediction based on the key supply chain conversation flow data is performed; The performing abnormal behavior pattern analysis on the first supply chain conversation flow data to generate a first abnormal behavior pattern analysis result includes: Performing feature extraction on the first supply chain conversation flow data to generate a supply chain conversation path vector graph of the first supply chain conversation flow data; Acquire interaction trend characteristic parameters between the data monitoring port and the supply chain conversation element included in the first supply chain conversation flow data; Perform abnormal behavior pattern analysis based on the supply chain conversation path vector diagram and the interaction trend characteristic parameters to generate a first abnormal behavior pattern analysis result; The step of obtaining the interaction trend characteristic parameters between the data monitoring port and the supply chain conversation elements included in the first supply chain conversation flow data includes: Analyzing an activity pattern of each supply chain conversation element in the first supply chain conversation flow data, and extracting time series data related to the supply chain conversation element based on the activity pattern of each supply chain conversation element; identifying respective key interaction pattern data from the time series data associated with the supply chain conversation elements; Calculate the interaction trend characteristic parameters between each supply chain conversation element and the data monitoring port based on each key interaction mode data, wherein the interaction trend characteristic parameters include at least one of interaction frequency, interaction duration, and interaction stability; The first supply chain conversation flow data includes supply chain order conversation flow data and supply chain product conversation flow data monitored for the same supply chain conversation element; extracting features from the first supply chain conversation flow data to generate a supply chain conversation path vector diagram of the first supply chain conversation flow data includes: Performing feature extraction on the supply chain order conversation flow data and the supply chain product conversation flow data respectively, and generating a supply chain order conversation path vector graph of the supply chain order conversation flow data and a supply chain product conversation path vector graph of the supply chain product conversation flow data; The abnormal behavior pattern analysis is performed according to the supply chain conversation path vector diagram and the interaction trend characteristic parameter to generate a first abnormal behavior pattern analysis result, including: Fusion of the supply chain order conversation path vector graph, the supply chain product conversation path vector graph, and the interactive trend vector of the interactive trend feature parameter to generate target fusion vector data; Performing abnormal behavior pattern analysis based on the target fusion vector data to generate a first abnormal behavior pattern analysis result; The step of fusing the supply chain order conversation path vector diagram, the supply chain product conversation path vector diagram and the interaction trend vector of the interaction trend feature parameter to generate target fusion vector data includes: Extracting a first key feature from the supply chain order session path vector diagram, wherein the first key feature includes a path feature, a length feature, and a direction feature of the supply chain order; Extracting a second key feature from the supply chain product conversation path vector graph, wherein the second key feature includes a flow path and a transaction volume change of the supply chain product; Representing the interaction trend vector of the interaction trend characteristic parameter as a numerical vector sequence, wherein the numerical vector sequence reflects the interaction intensity and trend between the data monitoring port and the supply chain conversation element; Fusing the first key feature, the second key feature and the numerical vector sequence to generate target fused vector data; The abnormal behavior pattern analysis of the first supply chain conversation flow data is performed to generate a first abnormal behavior pattern analysis result which is executed according to the abnormal behavior pattern prediction network; the abnormal behavior pattern analysis of the second supply chain conversation flow data is performed to generate a second abnormal behavior pattern analysis result which is executed according to the abnormal behavior pattern prediction network; the training step of the abnormal behavior pattern prediction network includes: Acquire sample supply chain conversation flow data and sample interaction trend characteristic parameters between the data monitoring port and the supply chain conversation elements included in the sample supply chain conversation flow data; the sample supply chain conversation flow data and the sample interaction trend characteristic parameters have abnormal behavior pattern annotation data; Extracting features of the example supply chain conversation flow data according to the initialized abnormal behavior pattern prediction network to generate an example supply chain conversation path vector graph of the example supply chain conversation flow data; According to the initialized abnormal behavior pattern prediction network, abnormal behavior pattern analysis is performed according to the example supply chain session path vector diagram and the example interaction trend characteristic parameters to generate an example abnormal behavior pattern analysis result; According to the example abnormal behavior pattern analysis result and the abnormal behavior pattern annotation data, the network parameters of the initialized abnormal behavior pattern prediction network are optimized until the network convergence condition is met, and the abnormal behavior pattern prediction network that meets the network convergence condition is generated; The example supply chain conversation flow data includes example supply chain order conversation flow data and example supply chain product conversation flow data; the example supply chain conversation flow data is feature extracted based on the initialized abnormal behavior pattern prediction network to generate an example supply chain conversation path vector diagram of the example supply chain conversation flow data, including: According to the initialized abnormal behavior pattern prediction network, feature extraction is performed on the example supply chain order conversation flow data and the example supply chain product conversation flow data respectively to generate an example supply chain order conversation path vector diagram of the example supply chain order conversation flow data and an example supply chain product conversation path vector diagram of the example supply chain product conversation flow data; The abnormal behavior pattern prediction network initialized according to the example supply chain session path vector diagram and the example interaction trend feature parameters is used to perform abnormal behavior pattern analysis to generate an example abnormal behavior pattern analysis result, including: Performing feature mapping on the example interaction trend feature parameters according to the initialized abnormal behavior pattern prediction network to generate an example interaction trend vector; Merging the example supply chain order conversation path vector graph, the example supply chain product conversation path vector graph, and the example interaction trend vector to generate example target vector data; An abnormal behavior pattern analysis is performed based on the example target vector data to generate an example abnormal behavior pattern analysis result.

2. The user demand forecasting method based on the supply chain settlement service system according to claim 1 is characterized in that: The method further comprises: Obtaining a priori user demand forecast data for the supply chain settlement service event; Determining predicted feedback data based on the priori user demand prediction data; If the first abnormal behavior pattern analysis result reflects that the data monitoring port has abnormal behavior pattern data, executing a data window shielding process for the data monitoring port includes: If the first abnormal behavior pattern analysis result reflects that abnormal behavior pattern data exists in the data monitoring port, and the predicted feedback data meets the shielding execution requirement, a data window shielding process for the data monitoring port is executed.

3. The user demand forecasting method based on the supply chain settlement service system according to claim 1 is characterized in that: If the first abnormal behavior pattern analysis result reflects that the data monitoring port has abnormal behavior pattern data, executing a data window shielding process for the data monitoring port includes: If the first abnormal behavior pattern analysis result reflects that abnormal behavior pattern data exists in the data monitoring port, a data window shielding process is performed for the data monitoring port according to the data window shielding component of the supply chain settlement service event.

4. The user demand forecasting method based on the supply chain settlement service system according to claim 3 is characterized in that: The abnormal behavior pattern data at the data monitoring port includes illegal access attempt data; the data window shielding component according to the supply chain settlement service event performs a data window shielding process for the data monitoring port, including: The security protection strategy of the supply chain settlement service system is activated to perform data window shielding processing on the illegal access attempt at the data monitoring port through the security protection strategy.

5. The user demand forecasting method based on the supply chain settlement service system according to claim 1 is characterized in that: The method further comprises: If the second abnormal behavior pattern analysis result reflects that there is no abnormal behavior pattern data, determining that the abnormal behavior pattern data at the data monitoring port has completed data window shielding; Determine the active range of abnormal data based on the second abnormal behavior pattern analysis result. If the active range of abnormal data is smaller than the active range of abnormal data determined based on the first abnormal behavior pattern analysis result, determine that the abnormal behavior pattern data at the data monitoring port has completed data window shielding.

6. The user demand forecasting method based on the supply chain settlement service system according to claim 1 is characterized in that: The step of predicting user demand based on the key supply chain conversation flow data monitored from the data monitoring port includes: Execute instructions according to user demand prediction to obtain key supply chain conversation flow data monitored from the data monitoring port; Matching the key supply chain conversation flow data with predefined historical user demand data to predict user demand based on the key supply chain conversation flow data; The historical user demand data includes historical product demand trends and historical product preference characteristics obtained based on analysis of past supply chain conversation flow data; The performing data matching analysis on the key supply chain conversation flow data and the pre-stored historical user demand data to predict user demand based on the key supply chain conversation flow data includes: Extracting current product demand information and current user preference trend data from the key supply chain conversation flow data; Compare and analyze the current product demand information with historical product demand trends to generate product demand forecast results; Comparing the current user preference trend data with historical product preference characteristics to generate a user preference prediction result; The product demand prediction result and the user preference prediction result are combined to generate a final user demand prediction result.

7. A user demand forecasting system based on a supply chain settlement service system, characterized in that: The user demand forecasting system based on the supply chain settlement service system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the user demand forecasting method based on the supply chain settlement service system as described in any one of claims 1 to 6 above.

Images