Purchase management method based on intelligent supply chain management platform

Abstract

The invention relates to the technical field of information, and provides a purchase management method based on an intelligent supply chain management platform, and the method comprises the steps: obtaining material demand data submitted by all departments, carrying out the standardization of the timestamp and demand of the material demand data, and forming standardized data containing an auxiliary classification dimension, a time dimension and a quantity dimension; generating an initial purchase plan table according to the real-time demand data and a preset purchase management rule, and grouping the purchase data in the initial purchase plan table according to a preset grouping rule to obtain a grouped purchase plan table; and obtaining response time prediction data, comparing the response time prediction data with a preset response time threshold value, if the response time prediction data is greater than the threshold value, optimizing the purchasing behavior by adjusting the priority and resource allocation, and determining an optimized purchasing strategy. The standardization, transparency and high efficiency of the purchasing process are realized, and the digital transformation and intelligent upgrading of enterprise purchasing are effectively promoted.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a procurement management method based on an intelligent supply chain management platform. Background Art

[0002] Supply chain management, as a core area of modern enterprise operation, is of decisive significance for improving resource allocation efficiency, reducing operation costs, and enhancing market competitiveness. With the acceleration of globalization and digitalization, enterprises have put forward higher requirements for the efficiency and transparency of the supply chain. Especially in the procurement management link, it directly affects the timeliness and economy of material supply. However, many current supply chain management solutions have shown significant defects in practical applications. Traditional methods often rely on manual operations or scattered systems, resulting in low efficiency of the procurement process and prominent data silo problems, and unable to achieve inter-departmental collaboration and real-time demand understanding. In addition, supplier management mostly stays offline, lacking systematic online integration and data analysis support, making it difficult for enterprises to scientifically evaluate and optimize procurement decisions. These limitations point to the core challenges that need to be broken through in the field of supply chain management. First, the real-time and accuracy of material demand management are insufficient. Due to the lack of a unified platform to integrate the demands of various departments, the procurement plan often gets out of touch with the actual demand. Second, the digitization and collaboration of the procurement process are limited. Especially, order processing and supplier docking still rely on manual or semi-automated methods, with low efficiency and easy to make mistakes. Finally, the lack of data analysis capabilities makes it impossible for enterprises to extract valuable insights from a large amount of procurement data, hindering the intelligence of decision-making support. Without solving these technical factors, it directly leads to unique problems such as the blindness of procurement plan formulation, the inefficiency of expert review, and the difficulty of implementing long-term cooperation models. Therefore, how to build an integrated and intelligent supply chain management platform to achieve real-time control of material demand and data-driven decision-making support for the procurement process has become a key issue in improving the procurement management efficiency of enterprises. The solution to this problem needs to focus on the technical integration of demand management, process optimization, and data analysis to ensure the efficient operation of the entire procurement management chain. Summary of the Invention

[0003] The present invention provides a procurement management method based on an intelligent supply chain management platform, mainly including: Obtain the material demand data submitted by each department, standardize the timestamps and demand quantities of the material demand data, classify and organize the standardized material demand data according to department names and material types to form standardized data including auxiliary classification dimensions, time dimensions, and quantity dimensions, where the auxiliary classification dimensions include departments and material types; extract the single-order single-purchase quantity of each department from the standardized data and calculate the purchase quantity fluctuation range respectively, eliminate abnormal purchase orders in the standardized data according to the preset fluctuation threshold to obtain real-time demand data; generate an initial purchase plan according to the real-time demand data and the preset purchase management rules, group the purchase data in the initial purchase plan according to the preset grouping rules to obtain a grouped purchase plan; according to the order feature data in the grouped purchase plan, combined with the access identification and blacklist identification in the preset supplier qualification database, calculate the matching degree between the order feature data and the supplier feature data, and publicly send a quotation invitation to the preset number of suppliers with the highest matching degree, access identification, and no blacklist identification for each purchase order, judge the order status based on the confirmation receipt feedback by the supplier and track the order status to obtain a tracking log; extract order processing data and supplier response time from the tracking log, use the pre-trained ARIMA model to calculate the predicted supplier response time to obtain response time prediction data, where the supplier response time refers to the time interval from when the order sends a quotation invitation to the supplier to when the supplier sends a confirmation receipt, and the response time prediction data includes the response time of the supplier predicted for future orders; compare the response time prediction data with the preset response time threshold, if it is greater than the preset response time threshold, optimize the order purchase behavior by adjusting the order priority and resource allocation to determine the optimized order purchase strategy; execute the order purchase task through the optimized order purchase strategy, obtain platform operation data from the platform operation log, and calculate the collaboration index data according to the platform operation data, where the collaboration index data includes order response speed and matching accuracy; analyze the collaboration index data through the decision tree algorithm to determine the key influencing factors affecting the collaboration index data, obtain the procurement cycle duration data including the planned procurement time and actual arrival time in the historical procurement data, calculate the Pearson correlation coefficient and significance level between the key influencing factors and the procurement cycle duration data, and generate the final procurement decision data according to the key influencing factors and their corresponding Pearson correlation coefficient and significance level; extract configuration parameters from the final procurement decision data, and iteratively update the order purchase strategy and resource scheduling rules of the platform according to the configuration parameters to obtain the full-chain operation status data, where the configuration parameters refer to parameters including supplier priority weights, response time index weights, maximum order volume that a supplier can bear, and emergency order resource quota ratios.

[0004] The technical solution provided by the embodiment of the present invention may include the following beneficial effects: The present invention discloses a procurement management method based on an intelligent supply chain management platform. This method collects real-time material demand data from various departments and classifies and processes it using a clustering algorithm to determine the real-time demand pattern. According to the demand pattern, the present invention generates an initial procurement plan and has a dynamic adjustment mechanism to cope with demand changes. The procurement plan is distributed to suppliers through the platform, and the order status is monitored simultaneously. The present invention also uses a time series analysis algorithm to calculate process efficiency indicators and adjusts task priorities and resource allocations through an optimization algorithm. In addition, the present invention uses a decision tree algorithm to analyze the key influencing factors of procurement decisions and continuously iteratively updates the operating status of the entire chain. This comprehensive, efficient, and flexible solution realizes the standardization, transparency, and efficiency of the procurement process, effectively promoting the digital transformation and intelligent upgrade of enterprise procurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1 It is a flowchart of the procurement management method based on the intelligent supply chain management platform of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0006] The technical solutions in the embodiments of the present invention will be clearly and detailedly described below with reference to the accompanying drawings in the embodiments of the present invention. The described embodiments are only a part of the embodiments of the present invention.

[0007] As Figure 1 , the procurement management method based on the intelligent supply chain management platform in this embodiment may specifically include: Step S101, obtain the material demand data submitted by each department, perform standardization processing on the timestamps and quantities of the material demand data, and classify and organize the standardized material demand data according to department names and material types to form standardized data including auxiliary classification dimensions, time dimensions, and quantity dimensions, where the auxiliary classification dimensions include departments and material types.

[0008] Obtain the material demand data submitted by each department through a real-time acquisition system to obtain preliminary data including timestamps, quantities, departments, and material types. Perform standardization processing on the timestamps and quantities in the preliminary data according to preset standardization rules to generate demand records with time identifiers. Classify and organize the demand records with time identifiers according to department names and material types to obtain standardized data including auxiliary classification dimensions, time dimensions, and quantity dimensions.

[0009] The material demand data submitted by each department is obtained through a real-time acquisition system, and preliminary data including timestamp, demand quantity, department, material type, and group bid section labels is obtained. The platform conducts penetrative management of the enterprise procurement process, and all demand submission and approval processes are recorded online to ensure compliance with internal control requirements. Exemplarily, a certain enterprise's material management system collects the office supply demands submitted by each department every day. The data includes the submission time such as 2025-05-15-08:00:00, the demand quantity such as 100 A4 papers, the department such as the Administration Department, and the material type such as office supplies. Real-time acquisition can be achieved through an online form. After each department fills in the form, the data is automatically uploaded to the central database to ensure the timeliness and accuracy of the data. This process relies on timestamps to record the submission moment, facilitating subsequent traceability and analysis, and helping to quickly respond to demands. Standardize the timestamps and demand quantities in the preliminary data to generate demand records with time identifiers. For example, timestamps may appear in different formats due to system differences, such as 2025-05-15-08:00:00 or 15 / 05 / 2025 / 08:00. The standardization process unifies them into the ISO format such as 2025-05-15T08:00:00Z to ensure cross-system compatibility. The standardization of demand quantities unifies the units and standardizes the presentation format. In one embodiment, the demand quantities are standardized. The standardization process includes unifying the procurement units and combining similar materials. For example, the unit of A4 paper purchased by the Administration Department is "ream", and that of the Finance Department is "box". After standardization, they are unified into "ream". 1 box is equal to 10 reams, and 60 boxes purchased by the Finance Department are converted to 600 reams. After combining similar materials, the total demand for A4 paper is 710 reams. After unifying the units, standardize the presentation format. For example, "710 reams of A4 paper" is standardized to "710 units of A4 paper (unit: ream)" to avoid misunderstandings caused by unclear expressions. The demand records generated after standardization contain unified time identifiers and quantified demand values, facilitating subsequent classification and statistics, and improving data processing efficiency. Classify and organize the demand records according to the department name and material type to obtain standardized data including auxiliary classification dimensions, time dimensions, and quantity dimensions. Specifically, classification can be achieved through grouped queries in the database. For example, the Administration Department submitted 600 units of A4 paper (unit: ream) and 50 units of pen refills (unit: piece) on 2025-05-15, and the Finance Department submitted 110 units of A4 paper (unit: ream). After classification, a record table is formed with departments and material types as dimensions. Auxiliary classification dimensions such as material priority can be further refined. For example, A4 paper is marked as high priority, and pen refills are marked as low priority. The time dimension retains the daily demand snapshot, and the quantity dimension summarizes the total demand. This standardized data supports multi-dimensional analysis, such as counting the consumption of office supplies by department or analyzing demand fluctuations over time, facilitating managers to optimize inventory allocation. It should be noted that the standardized data after classification and organization can generate reports through visualization tools, such as a bar chart showing the demand quantity of A4 paper for each department, or a line chart reflecting the monthly demand trend.This type of analysis helps to identify the reasons for the sudden increase in the demand for A4 paper in the administrative department, such as the onboarding of new employees, and thus enables advance stockpiling. In terms of technical effects, real-time collection ensures data timeliness, standardized processing improves data consistency, and classified sorting enhances data analyzability, ultimately achieving the efficiency and accuracy of material management and reducing the risks of inventory backlog or shortage. This extended solution complements the core classified sorting and jointly supports the intelligent decision-making of material management.

[0010] Step S102: Extract the single-order single-purchase quantity of each department from the standardized data, calculate the fluctuation range of the purchase quantity respectively, and eliminate the abnormal purchase orders in the standardized data according to the preset fluctuation threshold to obtain the real-time demand data.

[0011] Extract the single-order single-purchase quantity of each department from the standardized data, calculate the standard deviation of each department based on the sequence composed of multiple single-purchase quantities of each department as the fluctuation range value, where the multiple single-order single-purchase quantities of each department include the current order single-purchase quantity and the historical order single-purchase quantity. If the fluctuation range value exceeds the preset fluctuation threshold, mark the current purchase order as an abnormal purchase order, eliminate the abnormal purchase order, and obtain the real-time demand data.

[0012] Specifically, by extracting the single purchase quantity of each department based on standardized data and calculating the fluctuation range value, abnormal purchase data can be effectively identified to ensure the accuracy of real-time demand data. Specifically, the standardized data includes dimensions such as department, material type, timestamp, and demand quantity. For example, the administrative department submitted an order for 100 units of A4 paper on May 15, 2025. The single purchase quantity of each single order of each department is extracted respectively, that is, the demand quantity of each order is obtained, forming a sequence. For example, the single purchase quantities of the administrative department's recent five purchases of A4 paper are 100, 120, 110, 115, and 300 units. The calculation of the standard deviation reflects the degree of fluctuation of the sequence. Assuming that the preset fluctuation threshold is 50 units, if the standard deviation of the A4 paper purchase sequence of the administrative department is 75.8 units, exceeding the threshold, it is marked as abnormal purchase data. The single purchase quantity of 300 units of the current order, being significantly higher, may be an outlier and needs to be excluded. In one embodiment, the identification of abnormal purchase data can be achieved through database screening. The platform automatically extracts the recent order data of a certain material type of a certain department, generates a purchase quantity sequence, calculates the standard deviation and compares it with the threshold. For example, the recent five purchase quantities of refills by the finance department are 50, 55, 60, 52, and 200 units, and the standard deviation is 58.4 units, exceeding the threshold of 50 units. The current order of 200 units is marked as abnormal. After exclusion, the remaining data forms the real-time demand data. For example, the real-time demand for refills by the finance department is calculated based on the average value of 50, 55, 60, and 52 units, obtaining a more stable demand trend. It can be understood that the special application method for abnormal purchase orders provides flexibility for management. For example, the abnormal order of 300 units of A4 paper by the administrative department may be due to the temporary demand for a large-scale meeting. The department can submit an application to explain the reason and make a separate purchase after approval. This mechanism avoids delays in urgent demands caused by abnormal exclusion. After excluding abnormalities from the real-time demand data, it more accurately reflects the daily demand. For example, the average demand for A4 paper by the administrative department is about 111 units, which is convenient for inventory optimization. In one embodiment, the setting of the fluctuation threshold can be dynamically adjusted in combination with historical data. For example, by analyzing the A4 paper purchase data of the administrative department in the past six months, the average value of the standard deviation is 40 units, and the threshold can be set to 1.5 times the standard deviation, that is, 60 units, to ensure the reasonableness of the threshold. The dynamic threshold adapts to the fluctuation characteristics of different departments and material types, improving the accuracy of abnormal identification. For example, the generation of real-time demand data supports subsequent analysis, such as summarizing the monthly demand by department or predicting the inventory consumption by material type. After excluding abnormal data, the demand for A4 paper by the administrative department is more stable, and the demand for the next month is predicted to be about 110 units, making the procurement plan more accurate. The special application mechanism ensures that temporary large demands are met. For example, the abnormal order of 200 units of refills by the finance department is used for new employee training after the application is confirmed. This solution jointly improves the adaptability and accuracy of material management through abnormal identification, data cleaning, and flexible approval.

[0013] Step S103: Generate an initial procurement plan based on real-time demand data and preset procurement management rules, and group the procurement data in the initial procurement plan according to the preset grouping rules to obtain a grouped procurement plan.

[0014] Match the material type, demand quantity, demand department, and timestamp fields in the real-time demand data with the preset procurement management rules to generate an initial procurement plan. The preset management rules include a framework procurement ratio allocation rule and a supplier capacity limit rule. The initial procurement plan includes the name of the procurement material, specification model, recommended procurement quantity, minimum / maximum procurement quantity limit, demand department, and planned procurement time. The recommended procurement quantity is calculated comprehensively based on the demand quantity, preset safety stock threshold, and supplier capacity. According to the preset grouping rules, aggregate or split the procurement data in the initial procurement plan to form multiple procurement groups. The grouping rules include classification by demand department, classification by material type, classification by supplier adaptation category, and grading by procurement urgency. Perform data standardization processing on each procurement group to generate a grouped procurement plan containing the group number, material category, set of demand departments, total procurement quantity, planned procurement time interval, and corresponding supplier range. The standardization processing includes unifying the procurement unit, merging similar materials, and marking the group leader or the docking supplier.

[0015] In one embodiment, the real-time demand data includes fields such as material type, demand quantity, demand department, timestamp, etc., which are matched with preset procurement management rules to generate an initial procurement plan. The procurement management rules include a framework procurement ratio allocation rule and a supplier capacity limit rule. For example, the framework procurement ratio allocation rule stipulates that for the A4 paper procurement volume of the Administration Department, 55% is provided by the primary supplier and 45% is provided by the backup supplier; the supplier capacity limit rule sets that the primary supplier can supply a maximum of 5,000 units of A4 paper per month. Based on the real-time demand data, the Administration Department needs 110 units of A4 paper on May 15, 2025. Combining with the safety stock threshold of 50 units, the recommended procurement quantity is 160 units. The initial procurement plan records the material name A4 paper, specification model 70g, recommended procurement quantity 160 units, minimum procurement quantity 100 units, maximum procurement quantity 200 units, demand department Administration Department, planned procurement time May 20, 2025. The supplier capacity ensures that the procurement volume does not exceed the supplier's upper limit, guaranteeing stable supply. For example, the initial procurement plan is aggregated or split according to preset grouping rules, and the grouping rules include classification by demand department, material type, supplier adaptation category, and procurement urgency. Both the Administration Department and the Finance Department purchase A4 paper, which can be aggregated by material type to generate an A4 paper procurement group; if the procurement urgencies are different, such as the Administration Department's procurement is regular and the Finance Department's procurement is urgent, then it is split according to the urgency. After grouping, the regular procurement group of A4 paper includes the Administration Department's demand of 110 units and the Finance Department's regular demand of 600 units, totaling 710 units. The grouping rules flexibly adapt to different scenarios to ensure a clear procurement plan. The grouped procurement plan records the group number A4-001, material category A4 paper, demand department set including the Administration Department and the Finance Department, total procurement quantity of 710 units of A4 paper (unit: ream), planned procurement time interval from May 20, 2025 to May 25, 2025, corresponding supplier range including primary supplier A and backup supplier B. The group leader is marked as Procurement Manager Zhang to ensure clear docking. It can be understood that the standardized grouped procurement plan provides clear guidance for subsequent procurement execution. For example, for 710 reams of A4 paper allocated according to the framework procurement ratio, primary supplier A supplies 391 reams and backup supplier B supplies 319 reams, which meets the supplier capacity. For the urgent procurement group, supplier B is given priority to shorten the delivery time. The grouped procurement plan supports summarizing demands by department or material type, facilitating procurement coordination and inventory management, and ensuring procurement efficiency and resource optimization.

[0016] Step S104: According to the order feature data in the grouped procurement plan table, combined with the access identification and blacklist identification in the preset supplier qualification library, calculate the matching degree between the order feature data and the supplier feature data. Send out a public quotation invitation for each purchase order to a preset number of suppliers with the highest matching degree, access identification and no blacklist identification. Judge the order status based on the confirmation receipt feedback by the supplier and track the order status to obtain the tracking log.

[0017] Extract the purchase order from the grouped procurement plan table to generate order feature data, where the order feature data includes material type, purchase quantity and order urgency. Based on the access identification and blacklist identification in the supplier qualification library, construct a supplier feature data set, where the access identification is determined according to the supplier's supply capacity score and on-time delivery rate, and the blacklist is determined according to the historical default records. Calculate the matching degree between the order feature data and the supplier feature data through the weighted Euclidean distance algorithm. Combined with the supplier's supply capacity, send out a public quotation invitation for each purchase order to a preset number of suppliers with the highest matching degree, access identification and no blacklist identification. Mark the order status according to the confirmation receipt feedback by the supplier, record the time when the order sends out the quotation invitation to the supplier, the supplier response time and the status change information, and generate a tracking log including the order basic information, supplier information and status track, where the confirmation receipt includes one of the three status information, and the three status information are confirmation, rejection and pending respectively.

[0018] Specifically, to extract purchase orders from the grouped procurement plan table to generate order feature data including material type, purchase quantity, order urgency level, and group section label, it is necessary to ensure that the data accurately reflects the procurement requirements. The order feature data includes material type, purchase quantity, and order urgency level. For example, based on the aforementioned grouped procurement plan table, the demand for A4 paper in group number A4-001 is 710 reams, the material type is A4 paper, the purchase quantity is 710 reams, and the urgency level is regular procurement. When extracting, the system automatically parses the fields of the plan table to generate structured order feature data, such as material type: A4 paper, purchase quantity: 710 reams, urgency level: regular. It should be noted that the urgency level can be determined according to the time requirements submitted by the demand department or the out-of-stock emergency status. For example, when the inventory is lower than the safety threshold of 50 reams, it is automatically marked as urgent. For example, when constructing the supplier feature data set, the supplier qualification library provides an access identifier and a blacklist identifier. The access identifier is based on the supply capacity score and the on-time delivery rate. For example, the supply capacity score of supplier A is 85 points, and the on-time delivery rate is 95%, so the access identifier is "high-quality"; supplier C is included in the blacklist due to historical default records, and the blacklist identifier is "restricted cooperation". Specifically, the supply capacity score comprehensively evaluates the supplier's production capacity, logistics capacity, and financial stability, and the on-time delivery rate counts the performance data in the past 12 months. The supplier feature data set thus includes fields such as supplier name, access identifier, and blacklist identifier to ensure that high-quality suppliers are preferentially selected during procurement. It can be understood that when constructing the supplier feature data set, it can also be based on multi-dimensional labels in the supplier qualification library (such as material type adaptation label, green supply chain label, human carrying capacity score). For example, supplier A is marked as a "high-priority emergency order supplier" and has a human carrying capacity score of 90 points (out of 100). When the system calculates the matching degree through the weighted Euclidean distance algorithm, it simultaneously checks the current human carrying capacity of the supplier (such as current order volume / maximum carrying capacity ≤ 80%). If it exceeds the threshold, it automatically triggers the procurement process for alternative suppliers. In one embodiment, the matching degree between the order feature data and the supplier feature data is calculated through the weighted Euclidean distance algorithm. The order feature data such as purchase quantity and urgency level is matched with the supplier's supply capacity and on-time delivery rate. The core logic is to first standardize the feature data of different dimensions to a unified dimension, then quantify the "difference degree" between the order and the supplier through the weighted Euclidean distance formula, and finally convert the distance into a matching degree. For example, for a regular order of 710 reams of A4 paper, first process the data and standardize the purchase quantity. Assuming the maximum purchase quantity is 1000 reams, 710 reams of A4 paper is standardized to 0.71 after standardization; the urgency level of a regular order is set to 1 (assuming an urgent order is 3 and a relatively urgent order is 2). The supply capacity score of supplier A is 4 (out of 5), and after standardization, it is =0.8, and the on-time delivery rate is 0.95 (a 100% on-time delivery rate is recorded as 1); the supply capacity score of supplier B is 3, and after standardization, it is = 0.6, and the on-time rate is 0.90. The calculation formula for the weighted Euclidean distance is: , where X is the order feature data vector, Y is the supplier feature data vector, w i is the weight of the i-th feature, n is the number of features, x i is the standardized value of the i-th feature of the order, y i is the standardized value of the i-th feature of the supplier. In this embodiment, the number of features n = 2, the order feature data vector X = [0.71, 1], and the feature data vector Y of supplier A A = [0.8, 0.95], and the feature data vector Y of supplier B B = [0.6, 0.90]. The weight of the supply capacity w1 = 0.6, and the weight of the on-time rate w2 = 0.4. Calculate the weighted Euclidean distance between supplier A and the order: ≈ 0.0765. Convert the weighted Euclidean distance into a matching degree. The formula is: Matching degree = , then the matching degree of supplier A is: ≈ 0.93. Calculate the weighted Euclidean distance between supplier B and the order: ≈ 0.1061, and the matching degree of supplier B is: ≈0.90. Supplier A has a high supply capacity score and excellent on-time rate, and the matching score is 0.93 after calculation; Supplier B scores 0.90. The system selects Supplier A, which has no blacklist flag and the highest matching degree. It should be noted that in the weight setting, the supply capacity accounts for 60% and the on-time rate accounts for 40% to balance supply stability and delivery efficiency. It can be understood that after determining the pre-set number of in-stock suppliers with the highest matching degree, access identification, and no blacklist flag that meet the procurement conditions, the system automatically sends out a quotation invitation and publicly pushes the quotation requirements to the pre-set number of in-stock suppliers that meet the procurement conditions. After receiving the invitation, the supplier can quote online through the platform. At the same time, the platform triggers the expert extraction module to automatically extract evaluation experts according to pre-set rules (such as professional matching degree, expert avoidance mechanism, random extraction algorithm) to ensure the fairness and impartiality of the evaluation process. In the evaluation process, the transaction result takes the price as the main decision factor and is comprehensively determined in combination with the comprehensive score of the supplier (such as supply capacity score, historical default records, delivery timeliness, etc.). Among them, the number of suppliers can be determined according to the procurement type. For example, only 5 suppliers are required for regular procurement such as daily consumables, and 15 suppliers are required for special procurement of major matters. Exemplarily, after the supplier returns the confirmation receipt, the system marks the order status and generates a tracking log. Supplier A responds "confirmed" to the order of 710 reams of A4 paper, and the order status is updated to "confirmed". The tracking log records the order number A4-001-01, material type A4 paper, Supplier A, the time of sending the quotation invitation 2025-05-20-10:00, the response time 2025-05-20-12:00, and the status is changed to "confirmed". Specifically, if the supplier responds "pending", the system will re-conduct an open tender for the second-best supplier after 24 hours. The log also includes the historical record of status changes, such as from "pending procurement" to "confirmed", which is convenient for traceability. It can be understood that the above process ensures efficient procurement of purchase orders, reasonable selection of suppliers, and clear status tracking. The extraction of order feature data improves the accuracy of demand matching, the screening of supplier feature datasets selects reliable partners, the calculation of matching degree optimizes resource allocation, and the tracking log provides data support for subsequent audits and optimizations.

[0019] Step S105: Extract the order processing data and the supplier response time from the tracking log, and use the pre-trained ARIMA model to calculate the predicted supplier response time to obtain the response time prediction data. Among them, the supplier response time refers to the time interval from when the order sends a quotation invitation to the supplier to when the supplier sends a confirmation receipt, and the response time prediction data includes the response time of the supplier predicted for future orders.

[0020] Calculate the actual response time based on the time of sending a quotation invitation to the supplier according to the order, the confirmation receipt time, and the corresponding supplier information, and exclude abnormal orders. Combine the historical response times of the suppliers to generate comprehensive response time data including timestamps, current response times, and historical response times. Use data preprocessing methods to clean the comprehensive response time data, remove outliers, and fill in missing values to obtain standardized time series data. Input the standardized time series data into a pre-trained ARIMA model to generate response time prediction data including predicted time points, supplier identifiers, and response time prediction values, where the ARIMA model is pre-trained with historical time series data composed of the historical response times of the suppliers.

[0021] For example, in the scenario of purchase order procurement and supplier response management, calculating the actual response time and generating comprehensive response time data are important steps in optimizing supplier management. The actual response time refers to the time difference from the time when a quotation invitation is sent to the supplier for an order to the time when the supplier confirms the receipt. For example, for order A4-002, a quotation invitation was sent to supplier B at 2025-05-20-09:00, and the confirmation receipt time was 2025-05-20-11:30, so the actual response time was 2.5 hours. When excluding abnormal orders, a threshold can be set. For example, if the response time exceeds 48 hours, it is regarded as abnormal. Suppose the response time of supplier C for order A4-003 is 72 hours, and the system automatically marks it as abnormal and excludes it. The comprehensive response time data is generated by combining the supplier's historical response time, and the historical response time is determined based on the average value in the past 12 months. For example, the historical response time of supplier B is 2.8 hours, and the response time of the current order A4-002 is 2.5 hours. The comprehensive data record is: timestamp 2025-05-20-11:30, current response time 2.5 hours, historical response time 2.8 hours. In one embodiment, data preprocessing cleans the comprehensive response time data. The box plot method can be used to remove outliers, and 1.5 times the interquartile range is set as the threshold. For example, the historical time series of supplier A is 2.0, 2.5, 3.0, 50.0 hours, and 50.0 hours is removed because it exceeds the threshold. The mean filling method can be used to fill in the missing values. Suppose the response time of a certain order of supplier D is missing, and the system fills it with its historical mean value of 3.2 hours. The standardized time series data obtained after cleaning has a unified format and no outliers. For example, the timestamp is accurate to minutes, the response time unit is hours, and it consists of a timestamp, the current response time, and the historical response time, such as: timestamp 2025-05-20-11:30, current response time 2.5 hours, historical response time 2.8 hours. Exemplarily, the ARIMA model training is based on the standardized time series data. The ARIMA model captures the trend and periodicity of the response time by analyzing the autoregressive, differencing, and moving average characteristics of the time series. During training, the system takes the identifier of supplier B, the prediction time point, and the historical time series (such as 2.5, 2.7, 2.4 hours) as inputs, and automatically adjusts the model parameters to generate a fitting model. After training is completed, the model can reflect the fluctuation law of the supplier's response time. In one embodiment, the future response time is predicted based on the trained ARIMA model. The system inputs the identifier of supplier B and the prediction time point, such as 2025-06-01-09:00, and the model outputs the predicted response time of 2.6 hours. The prediction data includes timestamp 2025-06-01-09:00, supplier B, and predicted response time 2.6 hours. The prediction results can guide the order procurement behavior and preferentially select suppliers with short response times.It should be noted that the above process optimizes the efficiency of supplier selection and order management through response time calculation, data cleaning, model training and prediction. The elimination of abnormal orders ensures data reliability, data preprocessing improves the quality of model input, and the ARIMA model prediction provides an accurate reference for future procurement behavior. This method realizes the reasonable allocation of time resources in procurement management and improves the response speed and stability of the supply chain.

[0022] Step S106: Compare the response time prediction data with a preset response time threshold. If it is greater than the preset response time threshold, optimize the order procurement behavior by adjusting the order priority and resource allocation, and determine the optimized order procurement strategy.

[0023] Compare the response time prediction data of each supplier with a preset response time threshold. If the response time is greater than the preset response time threshold, mark it as a non-compliant supplier. For orders won by non-compliant suppliers, increase the task priority according to the urgency. Reduce the quota of bidable orders for non-compliant suppliers and increase the quota for high-quality suppliers. Increase the index weight of on-time delivery rate in the order procurement process, recalculate the matching degree between the order and the supplier, and determine the optimized order procurement strategy.

[0024] For example, in the scenario of order procurement and supplier response management, the comparative analysis of supplier response time prediction data is a key link in optimizing supply chain management. The system compares the predicted response time with a preset threshold to screen out suppliers that exceed the standard. Suppose the preset response time threshold is 4 hours, the predicted response time of Supplier A is 3.5 hours, and that of Supplier B is 5 hours. Then Supplier B is marked as a supplier that exceeds the standard. The setting of the threshold is based on historical data analysis and usually takes the upper quartile of the supplier response time to ensure rationality. After marking the suppliers that exceed the standard, the system can generate an exception report, recording the time stamp when Supplier B exceeded the standard, such as 2025-06-01-09:00, and the predicted value of 5 hours, which is convenient for subsequent management decisions. In one embodiment, for an order in which a supplier that exceeds the standard wins the bid, the system adjusts the order priority according to the order urgency. The order urgency can be determined by the delivery deadline and customer priority. For example, Order A4-005 needs to be delivered before 2025-06-03, and the customer rating is high. After Supplier B, which exceeds the standard, wins the bid for the tender, the system automatically upgrades its priority to "urgent" and notifies the logistics department to arrange transportation resources in advance. On the contrary, for non-urgent orders such as A4-006, the delivery deadline is relatively loose, and the priority remains unchanged. This dynamic adjustment ensures the timely processing of critical orders. Specifically, reducing the order quota of suppliers that exceed the standard is an important means of optimizing resource allocation. Suppose the monthly quota of Supplier B is 50 orders. Due to exceeding the response time standard, the system reduces its quota to 30, while increasing the quota of high-quality Supplier C from 50 to 70. The evaluation of high-quality suppliers is based on historical response time, such as 2.8 hours, and on-time delivery rate, such as 98%. After the quota adjustment, the system generates a new quota table, recording the adjustment time of 2025-06-01 and the quota changes of each supplier, ensuring the transparency and traceability of the procurement process. For example, in the process of order procurement, increasing the weight of the "on-time delivery rate" indicator can optimize supplier matching. The original matching model may set the weights of response time, cost, and on-time rate as 4:3:3, and now it is adjusted to 3:3:4, highlighting the importance of the on-time rate. Taking Order A4-007 as an example, the on-time rate of Supplier C is 98% and the response time is 2.8 hours, while the on-time rate of Supplier D is 90% and the response time is 3.0 hours. After adjusting the weights, the matching score of Supplier C is higher than that of Supplier D, and Supplier C is preferentially selected during the review. The matching degree is calculated based on the weighted average, comprehensively considering each indicator, ensuring that the procurement result better meets the requirements of supply chain efficiency. In one embodiment, the optimized order procurement strategy is further improved through real-time monitoring and feedback. The system records the time stamp of each invitation to quote sent to the supplier, such as 2025-06-02-10:00, the order number A4-008, the eligible Supplier C, and the predicted response time of 2.7 hours. If the actual response time deviates from the predicted value, the system automatically adjusts the weight or quota for the next procurement. For example, if the actual response time of eligible Supplier C is 3.5 hours, the system may reduce its matching score.This closed-loop mechanism ensures the continuous optimization of procurement strategies to adapt to the dynamic performance of suppliers. It should be noted that the implementation of the above methods relies on data-driven decision support. The marking, priority adjustment, and matching optimization of non-compliant suppliers together form a complete closed-loop order management. Each link is supported by specific data, such as timestamps, predicted values, on-time rates, etc., to ensure transparent and quantifiable operations. This method effectively balances the matching of supplier resources and order requirements, improving the overall response efficiency and stability of the supply chain.

[0025] Step S107, execute the order procurement task through the optimized order procurement strategy, obtain the platform operation data from the platform operation log, and calculate the synergy index data based on the platform operation data. Among them, the synergy index data includes the order response speed and the matching accuracy rate.

[0026] Execute the order procurement task through the optimized order procurement strategy, send a quotation invitation to the supplier for each order, and record the order procurement data in real time to obtain the platform operation log. Among them, the platform operation log includes the order number, the time of sending the quotation invitation to the supplier, the receipt time, and the actually successful bidding supplier. Obtain the operation data from the platform operation log. For the orders that have received the confirmation receipt normally, calculate the average interval between the time of sending the quotation invitation to the supplier and the receipt time as the order response speed. Compare the actually successful bidding suppliers in the platform operation log with the target suppliers preset in the strategy, and count the proportion of the number of orders with consistent matching as the matching accuracy rate. Generate the synergy index data including the response speed and the matching accuracy rate.

[0027] In one embodiment, the optimized order procurement strategy executes the procurement order procurement task through an automated system to ensure that the order is efficiently pushed to the supplier interface. The system transmits order information such as order number, procurement materials, and quantity to the supplier in real time through the API interface. For example, order B5-001 contains 100 parts and needs to be delivered before 2025-06-05. The system pushes it to the interface of supplier C at 2025-06-02-14:00. After the push, the system records the platform operation log, including order number B5-001, the time of sending the quotation invitation to the supplier 2025-06-02-14:00, the receipt time 2025-06-02-14:30, and the actually successful bidding supplier C. The log is stored in the form of a timestamp to ensure traceability. This recording method facilitates subsequent analysis of order processing efficiency. For example, extract the operation data from the platform operation log, filter the orders that have received confirmation receipts normally, and calculate the interval between the time of sending the quotation invitation to the supplier and the receipt time. Suppose the log shows that the time of sending the quotation invitation to the supplier for order B5-001 is 2025-06-02-14:00, and the receipt time is 2025-06-02-14:30, with an interval of 0.5 hours; the interval for order B5-002 is 0.4 hours. The system counts the intervals of 100 orders in the current month and obtains an average response speed of 0.45 hours. The response speed reflects the efficiency of the supplier in processing orders and provides data support for optimizing the procurement strategy. Specifically, compare the actually successful bidding suppliers in the log with the target suppliers preset in the strategy and calculate the prediction accuracy rate. Suppose the strategy presets that orders B5-001 and B5-002 are tendered to suppliers C - L. The log shows that it is consistent with the actually tendered suppliers. Count 100 orders in the current month, and among them, 90 orders have the same actual suppliers as the preset ones, with an accuracy rate of 90%. A high accuracy rate indicates that the system effectively executes the optimization strategy and reduces human intervention. It should be noted that orders with a low accuracy rate may be due to temporary adjustments by suppliers or system misjudgments, and further analysis of the timestamps and supplier data in the log is required. In one embodiment, the synergy index data is generated by combining the response speed and the matching accuracy rate. For example, the system generates a report showing a response speed of 0.45 hours and a matching accuracy rate of 90%. The synergy index is comprehensively evaluated through weighted calculation. Suppose the weight of the response speed is 0.6 and the weight of the accuracy rate is 0.4, and the synergy score is obtained as 0.78. The report is presented in the form of a chart to facilitate managers' intuitive understanding of the supply chain efficiency. Preferably, the system can adjust the strategy according to the synergy index, such as preferentially selecting suppliers with a fast response speed. For example, if the response speed of supplier D is 0.6 hours, which is lower than the average value, the system may reduce its quota and increase the orders for supplier C. The log records the adjustment time 2025-06-03-09:00 and the details of the quota change to ensure transparency. This closed-loop feedback mechanism optimizes procurement through real-time data analysis and improves the supply chain synergy efficiency.

[0028] In step S108, analyze the synergy index data through a decision tree algorithm to determine the key influencing factors affecting the synergy index data, obtain the procurement cycle duration data including the planned procurement time and the actual arrival time in the historical procurement data, calculate the Pearson correlation coefficient and the significance level between the key influencing factors and the procurement cycle duration data, and generate the final procurement decision data according to the key influencing factors and their corresponding Pearson correlation coefficients and significance levels.

[0029] Construct a feature matrix including fields such as order response speed and matching accuracy rate according to the synergy index data. Apply a decision tree algorithm to analyze the feature matrix and extract the key influencing factors with an importance score greater than 0.1. Extract the procurement cycle duration data from the historical procurement data. Calculate the Pearson correlation coefficient and the significance level between each key influencing factor and the procurement cycle duration data to form a factor - coefficient mapping table with the key influencing factors, Pearson correlation coefficients, and significance levels as the core fields, where the significance level is calculated by performing a hypothesis test on the Pearson correlation coefficient. Generate the final procurement decision data including the factor optimization direction and the decision threshold according to the correlation coefficients and significance levels in the factor - coefficient mapping table.

[0030] In one embodiment, when constructing the feature matrix, collaborative metric data such as order response speed and matching accuracy rate need to be extracted from the platform operation logs. The feature matrix takes orders as rows and fields as response speed, accuracy rate, and other relevant variables such as order volume and material type, forming a structured data table for subsequent analysis. When training the decision tree, label data needs to be provided to the model. In this embodiment, the label is defined as "procurement collaboration impact degree", which is calculated by combining expert scores with historical procurement data. For example, the procurement collaboration impact degree of a certain order = 0.4×(response speed / historical average response speed)+0.3×(matching accuracy rate / 100)+0.3×(order volume / historical maximum order volume). Input the feature matrix (response speed, accuracy rate, etc.) and label data into the decision tree model, and the model is trained by minimizing the mean squared error (MSE). For example, when analyzing the feature matrix using the decision tree algorithm, the system inputs the matrix into the model and calculates the importance scores of each field. The importance score is calculated by the decrease in node impurity: at each split, the model evaluates the proportion of impurity reduction brought by the split of this feature (such as the decrease in MSE), and the total contribution of this feature in all splits is its importance score. For example, the importance score of the response speed is 0.35, indicating that this feature contributes 35% to reducing the model prediction error. Suppose the analysis result shows that the importance score of the response speed is 0.35, the matching accuracy rate is 0.25, the order volume is 0.15, and the material type is 0.08. Screen the fields with scores greater than 0.1, and obtain the response speed, matching accuracy rate, and order volume as the key influencing factors. These factors reflect the core drivers of supply chain efficiency. Orders with a fast response speed can usually shorten the delivery cycle, high accuracy rate reduces error costs, and order volume affects procurement priorities. In one possible implementation, the procurement cycle duration data is extracted from historical procurement data. The system screens all orders in June 2025 and records the duration from order placement to delivery. For example, the procurement cycle of order B6-001 is 5 days, and that of order B6-002 is 7 days. The average cycle of 100 orders is 6 days. These data provide a basis for subsequent correlation analysis to ensure that the analysis results are close to the actual business. Specifically, when calculating the Pearson correlation coefficient between the key influencing factors and the procurement cycle duration, the system analyzes the response speed, matching accuracy rate, and order volume with the procurement cycle respectively. Suppose the correlation coefficient between the response speed and the procurement cycle is -0.65, indicating that the faster the response speed, the shorter the cycle; the matching accuracy rate is 0.40, indicating that the higher the accuracy rate, the cycle is slightly shortened; the order volume is 0.20, with a weak influence. The significance level is determined through statistical tests. Suppose the significance level of the response speed is 0.01, the matching accuracy rate is 0.05, and the order volume is 0.10. Generate a factor - coefficient mapping table, including fields: key factor, correlation coefficient, significance level.For example, the response speed (-0.65, 0.01), the matching accuracy rate (0.40, 0.05), and the order volume (0.20, 0.10). This table intuitively shows the influence intensity and credibility of each factor on the procurement cycle. For example, when generating procurement decision data based on the factor - coefficient mapping table, the system sets the optimization direction and decision threshold based on the correlation coefficient and significance level. The correlation coefficient of the response speed is -0.65 and the significance is high. The optimization direction is to preferentially select suppliers with a response speed lower than 0.5 hours, and the decision threshold is 0.5 hours. The correlation coefficient of the matching accuracy rate is 0.40, the optimization direction is to ensure that the accuracy rate is higher than 90%, and the threshold is 90%. The influence of the order volume is weak, and the threshold is set that the single - order volume does not exceed 1000 pieces to avoid excessive concentration of resources. The final decision data is output in tabular form, including factors, optimization directions, and thresholds, guiding order bidding. For example, the system adjusts the strategy and preferentially bids the order to supplier C with a fast response speed, records the adjustment time as 2025 - 07 - 02 - 09:00, ensuring the improvement of supply chain efficiency. It can be understood that when analyzing the feature matrix using the decision tree algorithm, the system can retrieve the historical review data of experts in related fields from the group expert database. For example, invite experts in the procurement field to verify the correlation analysis results between "order response speed" and "procurement cycle". The experts can adjust the threshold of the feature importance score based on industry experience (such as adjusting "importance score greater than 0.1" to 0.15). When calculating the Pearson correlation coefficient, the experts can set the industry standard of the significance level (such as p < 0.03) to ensure that the analysis results conform to the actual group procurement situation.

[0031] Step S109, extract configuration parameters from the final procurement decision data, and iteratively update the order procurement strategy and resource scheduling rules of the platform according to the configuration parameters to obtain the full - chain operation status data, where the configuration parameters refer to parameters including supplier priority weight, response time index weight, maximum order volume that the supplier can bear, and the proportion of emergency order resource quota.

[0032] Extract configuration parameters from the final procurement decision data, where the configuration parameters refer to parameters including supplier priority weight, response time index weight, maximum order volume that the supplier can bear, and the proportion of emergency order resource quota. According to the configuration parameters, update the order allocation algorithm built into the platform. Obtain the platform operation data after updating the order allocation algorithm, and generate full - chain operation status data including strategy version, update time, and index changes.

[0033] For example, in a supply chain management platform, extracting configuration parameters from the final procurement decision data is a crucial step in optimizing order procurement. The configuration parameters include supplier priority weights, response time metric weights, maximum order volume that a supplier can handle, and the proportion of emergency order resource quotas. These parameters directly affect the efficiency of the order allocation algorithm. The supplier priority weight reflects the importance of a supplier in procurement and is determined based on historical performance such as response speed and accuracy. For example, Supplier A has a weight of 0.8 due to its fast response speed and high accuracy, while Supplier B has a weight of 0.5. The response time metric weight is used to balance speed and cost. Assuming the platform prioritizes response speed, the weight is set to 0.6. The maximum order volume that a supplier can handle limits the order concentration of a single supplier. For example, the upper limit for Supplier A is 800 pieces. The proportion of emergency order resource quotas reserves resources for critical orders and is set at 20% to handle sudden demands. In a possible implementation, when updating the order procurement algorithm, the platform adjusts the logic according to the configuration parameters. For example, the algorithm preferentially selects suppliers with a weight higher than 0.7 while ensuring that the response time is less than 0.5 hours. The system sets the upper limit of the order volume for Supplier A at 800 pieces, and the excess is put out to tender to other eligible suppliers. Emergency orders are preferentially pushed to suppliers with fast responses according to the 20% quota. This update ensures that the algorithm fits the business requirements and improves procurement efficiency. For example, when obtaining the operation data of the updated platform, the system records the strategy version, update time, and metric changes. The full-chain operation status data reflects the algorithm's effectiveness. For example, the strategy version V2.0 was updated on 2025-07-03-10:00, the response speed decreased from 0.4 hours to 0.3 hours, and the matching accuracy increased from 92% to 95%. The data also includes the order completion rate, the proportion of delayed orders, etc., demonstrating the improvement in supply chain efficiency. These data provide a basis for subsequent optimization to ensure continuous improvement of decision-making. It can be understood that the above design of configuration parameter and algorithm update closely combines key factors in historical procurement data such as response speed and accuracy to ensure scientific and reasonable parameter settings. The comprehensive recording of operation status data provides the platform with a real-time monitoring and feedback mechanism, which helps to dynamically adjust strategies to adapt to changes in market demands.

[0034] Step S110: Send the final procurement result to the winning supplier in the form of a letter of acceptance, combine the final procurement result with the internally preset performance system to complete subsequent contract signing and performance tracking, and the tracking results form supplier big data for supplier portrait evaluation.

[0035] Exemplarily, the final procurement result is sent to the winning supplier in the form of a notice of successful transaction. After the procurement success result is synchronized to the internally preset performance system, the contract signing process will be automatically triggered, and the entire process of performance (such as delivery progress, quality acceptance) will be tracked. The platform supports the function of contract performance evaluation, collects data such as delivery on-time rate and after-sales service quality, and quantitatively evaluates the suppliers. The evaluation data is used to construct supplier portraits (such as performance ability scores, risk level labels), form supplier big data, and provide data support for subsequent supplier selection and procurement strategy optimization.

[0036] Only some preferred embodiments of the present invention are listed above, but the present invention is not limited thereto, and many improvements and transformations can be made. As long as the improvements and transformations are made on the basis of the basic principles of the present invention, they should be regarded as falling within the protection scope of the present invention.

Claims

1. A procurement management method based on an intelligent supply chain management platform, characterized in that, The method includes: Obtain the material demand data submitted by each department, standardize the timestamps and demand quantities of the material demand data, classify and organize the standardized material demand data according to department names and material types to form standardized data including auxiliary classification dimensions, time dimensions, and quantity dimensions, where the auxiliary classification dimensions include departments and material types; extract the single-order single-purchase quantity of each department from the standardized data and calculate the purchase quantity fluctuation range respectively, eliminate abnormal purchase orders in the standardized data according to a preset fluctuation threshold to obtain real-time demand data; generate an initial purchase plan according to the real-time demand data and a preset purchase management rule, group the purchase data in the initial purchase plan according to a preset grouping rule to obtain a grouped purchase plan; according to the order characteristic data in the grouped purchase plan, combine the access identification and blacklist identification in a preset supplier qualification library, calculate the matching degree between the order characteristic data and the supplier characteristic data, send a public quotation invitation to a preset number of suppliers with the highest matching degree, access identification, and no blacklist identification for each purchase order respectively, judge the order status according to the confirmation receipt feedback by the supplier and track the order status to obtain a tracking log; extract order processing data and supplier response time from the tracking log, calculate the predicted supplier response time using a pre-trained ARIMA model to obtain response time prediction data, where the supplier response time refers to the time interval from when the order sends a quotation invitation to the supplier to when the supplier sends a confirmation receipt, and the response time prediction data includes the predicted supplier response time for future orders; compare the response time prediction data with a preset response time threshold, if it is greater than the preset response time threshold, optimize the order purchase behavior by adjusting the order priority and resource allocation to determine the optimized order purchase strategy; execute the order purchase task according to the optimized order purchase strategy, obtain platform operation data from the platform operation log, calculate collaborative index data according to the platform operation data, where the collaborative index data includes order response speed and matching accuracy; analyze the collaborative index data through a decision tree algorithm to determine the key influencing factors affecting the collaborative index data, obtain the procurement cycle duration data including the planned procurement time and actual arrival time in the historical procurement data, calculate the Pearson correlation coefficient and significance level between the key influencing factors and the procurement cycle duration data, and generate final procurement decision data according to the key influencing factors and their corresponding Pearson correlation coefficient and significance level; extract configuration parameters from the final procurement decision data, and iteratively update the order purchase strategy and resource scheduling rules of the platform according to the configuration parameters to obtain full-chain operation status data, where the configuration parameters refer to parameters including supplier priority weights, response time index weights, maximum order quantity that a supplier can bear, and emergency order resource quota ratios.

2. The method according to claim 1, wherein The material demand data submitted by each department is obtained, the timestamp and demand quantity of the material demand data are standardized, and the standardized material demand data is classified and sorted according to department name and material type to form standardized data including auxiliary classification dimensions, time dimension and quantity dimension, wherein the auxiliary classification dimensions include department and material type, including: The material demand data submitted by each department is obtained through the real-time collection system, and preliminary data including timestamp, demand quantity, department and material type is obtained; The timestamps and demand quantities in the preliminary data are standardized according to the preset standardization rules to generate demand records with time stamps; Classify and organize demand records with time stamps according to department name and material type to obtain standardized data including auxiliary classification dimension, time dimension and quantity dimension.

3. The method according to claim 1, characterized in that, The process extracts the single purchase quantity of each department's single order from the standardized data and calculates the purchase quantity fluctuation range respectively. Abnormal purchase orders in the standardized data are eliminated according to a preset fluctuation threshold to obtain real-time demand data, including: Extract the single purchase quantity of each department's single order from the standardized data, and calculate the standard deviation of each department based on the sequence of multiple single purchase quantities of each department as the fluctuation range value. The single purchase quantity of multiple single orders of each department includes the single purchase quantity of the current order and the single purchase quantity of historical orders. If the fluctuation range value exceeds the preset fluctuation threshold, the current purchase order will be marked as an abnormal purchase order, and the abnormal purchase order will be eliminated to obtain real-time demand data.

4. The method according to claim 1, characterized in that, The method of generating an initial procurement plan based on real-time demand data and preset procurement management rules, and grouping procurement data in the initial procurement plan based on preset grouping rules to obtain a grouped procurement plan includes: Matching the material type, demand quantity, demand department, and timestamp fields in the real-time demand data with preset procurement management rules to generate an initial procurement plan, wherein the preset management rules include framework procurement ratio allocation rules and supplier carrying capacity limitation rules. The initial procurement plan includes the name of the purchased material, specification model, recommended purchase quantity, minimum / maximum purchase quantity limit, demand department, and planned procurement time. The recommended purchase quantity is calculated based on the demand quantity, the preset safety stock threshold, and the supplier carrying capacity. Aggregate or split the procurement data in the initial procurement plan table according to preset grouping rules to form multiple procurement groups, wherein the grouping rules include classification by demand department, classification by material type, classification by supplier adaptation category, and classification by procurement urgency; Data standardization is performed on each procurement group to generate a group procurement plan that includes group number, material category, demand department set, procurement quantity summary, planned procurement time range and corresponding supplier range. Among them, standardization includes unifying procurement units, merging similar materials, and marking group leaders or docking suppliers.

5. The method according to claim 1, characterized in that Based on the order feature data in the grouped procurement plan, combined with the access identification and blacklist identification in the preset supplier qualification database, calculate the matching degree between the order feature data and the supplier feature data, and send out public quotation invitations for each purchase order to a preset number of suppliers with the highest matching degree, access identification and no blacklist identification. Judge the order status according to the confirmation receipt feedback by the supplier and track the order status to obtain a tracking log, including: Extract the purchase order from the grouped procurement plan to generate order feature data, where the order feature data includes material type, purchase quantity, and order urgency; Based on the access identification and blacklist identification in the supplier qualification database, construct a supplier feature data set, where the access identification is determined according to the supplier's supply capacity and delivery on-time rate, and the blacklist is determined according to historical default records; Calculate the matching degree between the order feature data and the supplier feature data through the weighted Euclidean distance algorithm, and combined with the supplier's supply capacity, send out public quotation invitations for each purchase order to a preset number of suppliers with the highest matching degree, access identification and no blacklist identification; Mark the order status according to the confirmation receipt feedback by the supplier, record the time when the order sends out the quotation invitation to the supplier, the supplier response time, and the status change information, and generate a tracking log including order basic information, supplier information, and status track. Among them, the confirmation receipt includes one of the three status information, and the three status information are confirmation, rejection, and pending.

6. The method according to claim 1, wherein Extract the order processing data and the supplier response time from the tracking log, and use the pre-trained ARIMA model to calculate the predicted supplier response time to obtain the response time prediction data. Among them, the supplier response time refers to the time interval from when the order sends out the quotation invitation to the supplier to when the supplier sends the confirmation receipt, and the response time prediction data includes the predicted response time of the supplier for future orders, including: Calculate the actual response time according to the time when the order sends out the quotation invitation to the supplier, the confirmation receipt time, and the corresponding supplier information, and eliminate abnormal orders, and generate comprehensive response time data including timestamp, current response time, and historical response time in combination with the supplier's historical response time; Use data preprocessing methods to clean the comprehensive response time data, remove outliers and fill in missing values to obtain standardized time series data; Input the standardized time series data into the pre-trained ARIMA model to generate response time prediction data including predicted time points, supplier identification, and response time prediction values. Among them, the ARIMA model is pre-trained by the historical time series data composed of the supplier's historical response time.

7. The method according to claim 1, characterized in that, Compare the response time prediction data with the preset response time threshold. If it is greater than the preset response time threshold, optimize the order procurement behavior by adjusting the order priority and resource allocation to determine the optimized order procurement strategy, including: Compare the response time prediction data of each supplier with the preset response time threshold. If the response time is greater than the preset response time threshold, mark it as an over-standard supplier; For orders where the bid of a substandard supplier is successful, increase the task priority according to the urgency level; Reduce the bid quota for substandard suppliers and increase the quota for high-quality suppliers; Increase the index weight of the on-time delivery rate during the order procurement process, recalculate the matching degree between the order and the supplier, and determine the optimized order procurement strategy.

8. The method according to claim 1, wherein Execute the order procurement task through the optimized order procurement strategy, obtain the platform operation data from the platform operation log, and calculate the collaboration index data based on the platform operation data. Among them, the collaboration index data includes the order response speed and the matching accuracy rate, including: Execute the order procurement task through the optimized order procurement strategy, send a quotation invitation to the supplier for each order, and record the order procurement data in real time to obtain the platform operation log. Among them, the platform operation log includes the order number, the time of sending the quotation invitation to the supplier, the receipt time, and the actually successful bidder; Obtain the operation data from the platform operation log, and for the orders that have received the confirmation receipt normally, calculate the average interval between the time of sending the quotation invitation to the supplier and the receipt time as the order response speed; Compare the actually successful bidders in the platform operation log with the target suppliers preset in the strategy, and count the proportion of the number of orders with consistent matches as the matching accuracy rate; Generate collaboration index data including the response speed and the matching accuracy rate.

9. The method according to claim 1, characterized in that, Analyze the collaboration index data through the decision tree algorithm to determine the key influencing factors affecting the collaboration index data, obtain the procurement cycle duration data including the planned procurement time and the actual arrival time in the historical procurement data, calculate the Pearson correlation coefficient and the significance level between the key influencing factors and the procurement cycle duration data, and generate the final procurement decision data according to the key influencing factors and their corresponding Pearson correlation coefficients and significance levels, including: Construct a feature matrix including fields such as the order response speed and the matching accuracy rate according to the collaboration index data; Apply the decision tree algorithm to analyze the feature matrix and extract the key influencing factors with an importance score greater than 0.1; Extract the procurement cycle duration data from the historical procurement data; Calculate the Pearson correlation coefficient and the significance level between each key influencing factor and the procurement cycle duration data, and form a factor-coefficient mapping table with the key influencing factors, Pearson correlation coefficients, and significance levels as the core fields. Among them, the significance level is calculated through hypothesis testing of the Pearson correlation coefficient; Generate the final procurement decision data including the factor optimization direction and the decision threshold according to the correlation coefficients and significance levels in the factor-coefficient mapping table.

10. The method according to claim 1, wherein Extract the configuration parameters from the final procurement decision data, and iteratively update the order procurement strategy and resource scheduling rules of the platform according to the configuration parameters to obtain the full-chain operation status data. Among them, the configuration parameters refer to the parameters including the supplier priority weight, the response time index weight, the maximum order volume that the supplier can bear, and the emergency order resource quota ratio, including: Send the final procurement result to the winning supplier in the form of a notice of successful transaction, combine the final procurement result with the internally preset performance system to complete subsequent contract signing and performance tracking, and the tracking results form the supplier big data for supplier portrait evaluation.

Images