Supplier comprehensive management method and system based on real-time AI analysis

Abstract

The invention provides a supplier comprehensive management method and system based on real-time AI analysis, and relates to the technical field of supplier comprehensive management, and the method comprises the steps: extracting supplier core capability features from standardized supplier historical cooperation data based on a depth factorization model; dynamic performance features are extracted from the supplier real-time operation data through a time sequence attention mechanism; performing AI comprehensive evaluation on the dynamic performance characteristics based on the supplier core capability characteristics to obtain a supplier real-time rating result, and performing weight adjustment in combination with the purchase demand priority to generate a supplier grading directory; a supplier digital portrait system is constructed based on a supplier grading directory, cooperation adaptability clustering is performed through a density clustering algorithm, a supplier classification cooperation model is obtained, intelligent matching and cooperation strategy optimization are performed on purchase demands, a customized cooperation scheme is generated, dynamic iteration updating is performed in combination with real-time market dynamics, and a supplier classification cooperation model is obtained. And full-life-cycle comprehensive management of suppliers is realized.

Description

Technical Field

[0001] This invention relates to the field of supplier integrated management technology, specifically a supplier integrated management method and system based on real-time AI analysis. Background Technology

[0002] Currently, supplier management largely relies on static evaluations of historical cooperation data, which suffers from limitations such as a single evaluation dimension and insufficient real-time capability. It's difficult to comprehensively capture a supplier's core capabilities and performance, and the multi-source data (such as order execution, quality inspection, and logistics services) throughout the cooperation process are scattered and independent, lacking effective integration and real-time analysis mechanisms. This results in an inability to promptly perceive changes in supplier operational status. Furthermore, the evaluation results are not adequately adjusted dynamically based on procurement needs and market dynamics, leading to a lack of targeted supplier matching and cooperation strategies. This makes it difficult to adapt to complex and ever-changing procurement scenarios and hinders refined management of the supplier's entire lifecycle, impacting procurement efficiency and cooperation stability.

[0003] Therefore, a comprehensive supplier management method and system based on real-time AI analysis is provided. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide a supplier integrated management method and system based on real-time AI analysis.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a supplier comprehensive management method based on real-time AI analysis, the method comprising: The pre-collected historical cooperation data of suppliers is standardized, and the core capability characteristics of suppliers are extracted from the standardized historical cooperation data based on the deep factor decomposition model. Based on the multi-source real-time data acquisition terminal, multi-dimensional data collection and real-time verification are carried out on the entire process of supplier cooperation to obtain real-time operational data of suppliers, and dynamic performance characteristics are extracted from the real-time operational data of suppliers through the time-series attention mechanism. Based on the core capabilities of the suppliers, the dynamic performance characteristics are comprehensively evaluated by AI to obtain the real-time rating results of the suppliers. The weight of the real-time rating results of the suppliers is adjusted in combination with the priority of procurement needs to generate a supplier classification list. Based on the supplier classification directory, a supplier digital profile system is constructed, and the suppliers in the profile system are subjected to cooperation adaptability clustering using a density clustering algorithm to obtain a supplier classification cooperation model. Based on the supplier classification and cooperation model, procurement needs are intelligently matched and cooperation strategies are optimized to generate customized cooperation solutions. These customized cooperation solutions are then dynamically iterated and updated in conjunction with real-time market dynamics, enabling comprehensive management of the entire supplier lifecycle.

[0006] Furthermore, the process of extracting core capability features of suppliers from standardized historical supplier cooperation data based on a deep factor decomposition model includes: Multi-dimensional historical feature data is extracted from standardized supplier historical cooperation data, and the multi-dimensional historical feature data is discretized and normalized based on a preset feature dimension system to obtain the supplier basic feature set. First-order feature filtering is performed on the supplier basic feature set to obtain an effective basic feature set; cross feature data is extracted from the multi-dimensional historical feature data, and feature embedding is performed on the cross feature data to obtain the supplier cross feature set; The effective basic feature set and the supplier cross feature set are input into the deep factorization model to obtain the first-order feature interaction vector and the higher-order feature interaction vector. The attention weight distribution of the first-order feature interaction vector and the higher-order feature interaction vector is calculated based on the feature interaction attention layer to obtain the feature interaction weight. The first-order feature interaction vector and the higher-order feature interaction vector are weighted and fused based on the feature interaction weights to obtain a fused feature vector; the fused feature vector is then subjected to feature dimensionality reduction to obtain the supplier's core capability features.

[0007] Furthermore, the process of obtaining real-time operational data from suppliers by collecting and verifying multi-dimensional data across the entire supplier cooperation process using multi-source real-time data acquisition terminals includes: The system obtains real-time order data by acquiring supplier order execution data through the order management system interface and adding timestamps; it also obtains real-time quality data by acquiring process quality data and batch sampling data during product production through the quality inspection equipment data acquisition interface and adding timestamps; it obtains real-time logistics data by acquiring cargo transportation data through the logistics GPS tracking interface and adding timestamps; it obtains real-time service data by acquiring service interaction data through the after-sales service platform interface and adding timestamps; and it obtains real-time financial data by acquiring financial interaction data through the financial settlement system interface and adding timestamps. The real-time order data, real-time quality data, real-time logistics data, real-time service data, and real-time financial data are aligned with time windows and correlated to obtain real-time multi-source fused data; based on preset verification rules, the real-time multi-source fused data is cross-validated to obtain real-time supplier operation data.

[0008] Furthermore, the process of extracting dynamic fulfillment features from supplier real-time operational data through a temporal attention mechanism includes: Based on the temporal attention mechanism, real-time dynamic data is extracted from the supplier's real-time operation data, and the real-time dynamic data is sorted in chronological order to obtain a real-time temporal data sequence; local temporal features are extracted from the real-time temporal data sequence using a temporal attention convolutional layer to obtain local temporal features. The attention weights of each time step in the local temporal features are calculated based on the temporal attention encoding layer to obtain weighted temporal features; global average pooling is then performed on the weighted temporal features to obtain dynamic performance features.

[0009] Furthermore, the process of conducting a comprehensive AI evaluation of dynamic performance characteristics based on the aforementioned core supplier capabilities to obtain real-time supplier rating results includes: The supplier's core capability features are concatenated with the corresponding dynamic performance features to obtain a comprehensive evaluation feature vector; the comprehensive evaluation feature vector is input into a pre-trained supplier comprehensive evaluation model to obtain an initial evaluation score; the initial evaluation score is classified into levels based on a preset rating threshold to obtain the supplier's real-time rating result.

[0010] Furthermore, the training process of the supplier comprehensive evaluation model includes: Obtain the core capability characteristics of suppliers and the corresponding dynamic performance characteristics of the period from several sets of historical cooperation data collected over historical periods; and use cooperation satisfaction as the tag value; The core capability characteristics of suppliers, their corresponding dynamic performance characteristics, and tag values ​​from several sets of historical cooperation data collected over several periods are grouped and labeled, denoted as follows: It is a natural number; Will The core capabilities of suppliers, their dynamic performance characteristics for the corresponding period, and tag values ​​from historical cooperation data collected over a historical period are used as sample data. Less than The natural numbers, and using the sample data, the mean of the sample data is obtained, denoted as the sample set; The core capability characteristics of suppliers, the dynamic performance characteristics of the corresponding period, and the tag values ​​in the historical cooperation data of the remaining historical collection periods are used as the test set; a training sample set is formed based on the sample set and the test set. The GBDT model and the neural network model are trained based on the training sample set, and the trained model is denoted as the supplier comprehensive evaluation model.

[0011] Furthermore, the process of adjusting the weights of real-time supplier ratings based on procurement needs to generate a tiered supplier list includes: Obtain current procurement demand information and extract procurement demand priority features; calculate the weight coefficients of each procurement demand priority feature based on the analytic hierarchy process (AHP). The initial evaluation score of the supplier's real-time rating result is adjusted according to the weighting coefficients to obtain the adjusted evaluation score; The supplier grading list is generated by reclassifying the suppliers based on the adjusted evaluation scores and sorting them according to the grade order and the adjusted scores.

[0012] Furthermore, the process of constructing a supplier digital profile system based on the aforementioned supplier tiered directory includes: Extract the supplier name, rating level, adjusted evaluation score, core advantages, and suitable demand type of each supplier from the supplier classification list to construct a core capability profile; obtain the dynamic performance characteristics of the corresponding supplier to construct a dynamic performance profile; and extract semantic features from cooperation evaluation texts or complaint records based on natural language processing technology to construct a cooperation suitability profile. By integrating the core capability profile, dynamic performance profile, and cooperation adaptation profile of the corresponding supplier, a complete digital profile of the corresponding supplier is formed.

[0013] Furthermore, the process of obtaining a supplier classification cooperation model by performing cooperative adaptation clustering of suppliers in the profiling system using density clustering algorithms includes: Extract the cooperation adaptation feature set from the supplier digital profile system; based on Z-score standardization, perform feature standardization on the cooperation adaptation feature set to obtain the standardized adaptation feature set; Cluster analysis is performed on the standardized adaptation feature set based on the density clustering algorithm to obtain the corresponding clusters; feature analysis is performed on each cluster to extract the core features of the cluster, which include large-batch standardized supply clusters, small-batch customized supply clusters, and emergency order response clusters; By assigning cooperation scenario tags and matching procurement demand types to each cluster, a supplier classification cooperation model is obtained.

[0014] A second aspect of the present invention also provides a supplier integrated management system based on real-time AI analysis, comprising: a feature extraction module, a real-time acquisition module, a comprehensive evaluation module, a digital profiling module, and a dynamic cooperation module; The feature extraction module is used to standardize the pre-collected historical cooperation data of suppliers and extract the core capability features of suppliers from the standardized historical cooperation data based on the deep factor decomposition model. The real-time acquisition module collects and verifies multi-dimensional data throughout the supplier cooperation process based on a multi-source real-time data acquisition terminal, obtains real-time operational data of suppliers, and extracts dynamic performance characteristics from the real-time operational data of suppliers through a time-series attention mechanism. The comprehensive evaluation module performs an AI-based comprehensive evaluation of dynamic performance characteristics based on the core capability characteristics of the suppliers, obtains real-time supplier rating results, and adjusts the weight of the real-time supplier rating results in combination with the priority of procurement needs to generate a supplier tier list. The digital profiling module constructs a supplier digital profiling system based on the supplier hierarchical directory, and uses a density clustering algorithm to perform cooperation-adaptive clustering on the suppliers in the profiling system to obtain a supplier classification cooperation model. The dynamic cooperation module intelligently matches procurement needs and optimizes cooperation strategies based on the supplier classification cooperation model, generates customized cooperation solutions, and dynamically iterates and updates the cooperation solutions in conjunction with real-time market dynamics, thereby realizing comprehensive management of the entire supplier lifecycle.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: Through real-time AI analysis and multi-dimensional data fusion, it achieves precise, dynamic, and intelligent supplier management. Based on the core capability characteristics of suppliers extracted by a deep factor decomposition model, combined with dynamic performance characteristics captured by a temporal attention mechanism, the evaluation is grounded in both historical strength and real-time status. The tiered list generated by AI comprehensive evaluation and weighted adjustment of procurement demand priorities, along with the classification cooperation model constructed by density clustering, ensures the accuracy of supplier matching and the adaptability of cooperation strategies. Customized cooperation schemes and iterative updates driven by real-time market dynamics enable dynamic management from initial matching to the entire lifecycle of cooperation, effectively improving procurement efficiency, cooperation stability, and risk control capabilities, providing scientific support for enterprise procurement decisions. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a schematic diagram illustrating the steps of a comprehensive supplier management method based on real-time AI analysis.

[0018] Figure 2 This is a schematic diagram of a supplier integrated management system based on real-time AI analysis. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0020] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0021] Example 1: like Figure 1 As shown, a comprehensive supplier management method based on real-time AI analysis includes the following steps: Step S1: Standardize the pre-collected historical cooperation data of suppliers and extract the core capability features of suppliers from the standardized historical cooperation data based on the deep factor decomposition model. Step S2: Based on the multi-source real-time data acquisition terminal, multi-dimensional data collection and real-time verification are carried out on the entire process of supplier cooperation to obtain real-time operational data of suppliers, and dynamic performance characteristics are extracted from the real-time operational data of suppliers through the time-series attention mechanism. Step S3: Based on the core capability characteristics of the suppliers, perform a comprehensive AI evaluation of the dynamic performance characteristics to obtain the real-time rating results of the suppliers. Then, adjust the weight of the real-time rating results of the suppliers in combination with the priority of procurement needs to generate a supplier classification list. Step S4: Construct a supplier digital profile system based on the supplier classification list, and perform cooperation adaptability clustering on the suppliers in the profile system using density clustering algorithm to obtain a supplier classification cooperation model; Step S5: Based on the supplier classification and cooperation model, intelligently match procurement needs and optimize cooperation strategies to generate customized cooperation plans. Combine real-time market dynamics to dynamically iterate and update customized cooperation plans to achieve comprehensive management of the entire supplier lifecycle.

[0022] It should be further explained that, in the specific implementation process, the data standardization process for the pre-collected historical cooperation data of suppliers includes: It should be noted that the aforementioned supplier historical cooperation data refers to the supplier's full-process operational data recorded during past cooperation periods. This full-process operational data includes, but is not limited to, basic supplier information (e.g., supplier name, supplier qualification level, supplier registered address, and supplier business scope), cooperation performance data (e.g., order delivery time, delivery quantity accuracy rate, and number of delayed deliveries), product quality data (e.g., sampling pass rate, defective product return rate, and number of quality complaints), service support data (e.g., after-sales response time, problem resolution closure rate, and technical support satisfaction), and cost data (e.g., price fluctuation range, procurement cost ratio, and payment cycle suitability). Optionally, in this embodiment, the pre-collected supplier historical cooperation data undergoes data type unification (e.g., encoding text-based qualification data into numerical data and converting date-based data into time difference quantification values) and logical consistency verification (e.g., verifying the matching of delivery time with order cycle and the conformity of product quality data with sampling standards). The supplier historical cooperation data after logical consistency verification is then deduplicated (e.g., removing duplicate order fulfillment records) and outlier filtering (e.g., using box plotting to remove outliers). This application will obtain standardized historical cooperation data of suppliers by analyzing extreme price data or abnormal delays in delivery (within a certain range). The specific processing procedures will not be elaborated in this application.

[0023] It should be further explained that, in the specific implementation process, the process of extracting the core capability characteristics of suppliers from standardized historical cooperation data based on the deep factor decomposition model includes: It should be noted that the deep factorization model includes a factorization machine layer, namely the FM layer (Factorization Machine, or FM for short), a deep neural network layer, namely the DNN layer (Deep Neural Network, or DNN for short), and a feature interaction attention layer.

[0024] Optionally, in this embodiment of the application, multi-dimensional historical feature data is extracted from the standardized supplier historical cooperation data, and the multi-dimensional historical feature data is discretized and normalized based on a preset feature dimension system to obtain a supplier basic feature set. First-order feature filtering is performed on the supplier basic feature set to obtain an effective basic feature set; cross feature data is extracted from the multi-dimensional historical feature data, and feature embedding is performed on the cross feature data to obtain the supplier cross feature set; The effective basic feature set is input into the FM layer of the deep factorization model to obtain the first-order feature interaction vector; The supplier cross-feature set is input into the DNN layer of the deep factorization model, and high-order feature interaction learning is performed through a 3-layer fully connected network to obtain the high-order feature interaction vector. The attention weight distribution of the first-order feature interaction vector and the higher-order feature interaction vector is calculated based on the feature interaction attention layer to obtain the feature interaction weight. The first-order feature interaction vector and the higher-order feature interaction vector are weighted and fused based on the feature interaction weights to obtain a fused feature vector; the fused feature vector is then subjected to feature dimensionality reduction to obtain the supplier's core capability features.

[0025] It should be noted that the preset feature dimension system includes, but is not limited to, basic qualification dimensions, performance capability dimensions, quality control dimensions, service support dimensions, and cost adaptation dimensions. Specific feature dimensions can be expanded according to actual circumstances. Feature discretization refers to dividing continuous data (e.g., delivery time and quoted price) into several level intervals and encoding them. Feature normalization uses the Min-Max normalization method to map feature values ​​to the [0,1] interval, ensuring the comparability of features across different dimensions. Feature embedding refers to converting high-dimensional sparse cross-features into low-dimensional dense vectors through a shared embedding matrix. The FM layer is used to capture linear interaction relationships between features; the DNN layer is used to mine nonlinear complex interaction patterns between features, achieving interactive learning between features of different dimensions based on the FM and DNN layers; the feature interaction attention layer highlights the impact of key interaction features by calculating the contribution of each feature interaction item to the overall supplier evaluation.

[0026] Optionally, in this embodiment, the attention weight distribution of the first-order feature interaction vector and the higher-order feature interaction vector is calculated based on the feature interaction attention layer to obtain the feature interaction weights, specifically including: Based on the linear mapping layer, the first-order feature interaction vector and the higher-order feature interaction vector are dimensionally aligned to obtain a standard interaction feature matrix; based on the standard interaction feature matrix, attention query vector, key vector and value vector are constructed. The similarity between the query vector and the key vector is calculated based on the scaling dot product attention mechanism to obtain the original attention weights; the original attention weights are then subjected to Softmax normalization to obtain the initial interaction weights. The random forest algorithm is used to calculate the feature importance coefficient of each feature dimension in the multi-dimensional historical feature data, thereby obtaining the importance score of the corresponding feature dimension. The initial interaction weight is then weighted twice based on the importance score to obtain the feature interaction weight.

[0027] It should be noted that the linear mapping layer uses the ReLU activation function to address the vanishing gradient problem. The scaled dot product attention mechanism addresses gradient instability caused by the curse of dimensionality by dividing the inner product of the query vector and the key vector by the square root of the key vector's dimension. The calculation of the feature importance coefficient is based on optimizing the random forest model parameters using 10-fold cross-validation, i.e., setting the number of decision trees to 100 and the maximum depth to 10 to ensure the reliability of the feature importance coefficient; secondary weighting further reflects the interactive influence of core feature dimensions, improving the recognizability of core capability features.

[0028] To further explain, the deep factorization model enables multi-dimensional, full-level feature interaction learning, preserving the linear correlation of basic features and the non-linear relationship of complex cross features. It also highlights key influencing factors based on attention weights, effectively improving the representation accuracy of the supplier's core capability features and providing a reliable feature foundation for subsequent real-time evaluation.

[0029] It should be further explained that, in the specific implementation process, the process of obtaining real-time operational data of suppliers by collecting and verifying multi-dimensional data throughout the entire supplier cooperation process based on multi-source real-time data acquisition terminals includes: It should be noted that the multi-source real-time data acquisition terminal includes, but is not limited to, interfaces for order management systems, quality inspection equipment, logistics tracking, after-sales service platforms, and financial settlement systems. Real-time data transmission is achieved through RESTful APIs or MQTT protocols, and a collection frequency is set to ensure data timeliness. Then, corresponding real-time supplier operational data is collected according to the collection frequency. The supplier's real-time operational data refers to the supplier's dynamic operational data within the current cooperation period. This dynamic operational data includes, but is not limited to, order execution data (e.g., real-time production progress, delivered quantity, and remaining delivery period), real-time quality data (e.g., online testing data, batch sampling results, and process quality anomaly alarms), logistics tracking data (e.g., cargo location, transportation method, and estimated arrival time), service response data (e.g., after-sales consultation processing time, emergency response speed, and complaint handling progress), and financial interaction data (e.g., timeliness of invoice issuance and payment application response speed).

[0030] Optionally, in this embodiment, order execution data from suppliers is obtained through the order management system interface and timestamps are added to obtain real-time order data; process quality data and batch sampling data during product production are obtained through the quality inspection equipment data acquisition interface and timestamps are added to obtain real-time quality data; cargo transportation data is obtained through the logistics GPS tracking interface and timestamps are added to obtain real-time logistics data; service interaction data is obtained through the after-sales service platform interface and timestamps are added to obtain real-time service data; and financial interaction data is obtained through the financial settlement system interface and timestamps are added to obtain real-time financial data. The real-time order data, real-time quality data, real-time logistics data, real-time service data, and real-time financial data are aligned with time windows and correlated to obtain real-time multi-source fused data; based on preset verification rules, the real-time multi-source fused data is cross-validated to obtain real-time supplier operation data.

[0031] It should be noted that the preset verification rules include, but are not limited to, data consistency verification, logical rationality verification, and abnormal data verification; data consistency verification refers to comparing related data collected by different data acquisition terminals, for example, whether the delivered quantity in the order management system interface is consistent with the shipped quantity in the logistics tracking interface; logical rationality verification refers to verifying whether the data conforms to business logic, for example, the expected arrival time cannot be earlier than the shipment time; abnormal data verification refers to... Data exceeding the normal range is detected in principle; data that fails verification is corrected through manual verification to ensure the accuracy of the supplier's real-time operational data.

[0032] It should be further explained that, in the specific implementation process, the process of extracting dynamic performance characteristics from the supplier's real-time operational data through the time-series attention mechanism includes: It should be noted that the time-series attention mechanism is used to extract dynamic performance features from the supplier's real-time operation data in order to achieve a unified data dimension system between the supplier's real-time operation data and multi-dimensional historical feature data.

[0033] Optionally, in this embodiment of the application, based on the temporal attention mechanism, real-time dynamic data is extracted from the supplier's real-time operational data, and the real-time dynamic data is labeled with a status, thereby extracting dynamic performance features from the real-time dynamic data. The specific steps are as follows: The real-time dynamic data is sorted in chronological order to obtain a real-time time-series data sequence; local time-series features are extracted from the real-time time-series data sequence using a time-series attention convolutional layer to obtain local time-series features; It should be noted that the kernel size of the temporal attention convolutional layer is set to 3, and the stride is set to 1.

[0034] The attention weights of each time step in the local temporal features are calculated based on the temporal attention encoding layer to obtain weighted temporal features; global average pooling is then performed on the weighted temporal features to obtain dynamic performance features.

[0035] It should be noted that the temporal attention encoding layer assigns higher attention weights to recent data by calculating the similarity between features at each time step and features at the current time, thereby ensuring that features can reflect changes in the supplier's performance status in real time. For example, if the weight of real-time dynamic data corresponding to the last three time windows is twice that of historical data, then data marked as abnormal will receive an additional half of the attention weight.

[0036] It should be further explained that, in the specific implementation process, the process of conducting a comprehensive AI evaluation of dynamic performance characteristics based on the aforementioned core supplier capabilities to obtain the supplier's real-time rating results includes: It should be noted that existing supplier evaluation methods mostly adopt a static evaluation model with fixed indicator weights. This model cannot respond in real time to changes in supplier performance status and does not incorporate dynamic adjustments to procurement needs. Consequently, the evaluation results do not match actual cooperation requirements, affecting the scientific nature of procurement decisions. This application achieves real-time and practical evaluation results by combining AI-driven dynamic evaluation with demand-adaptive adjustments.

[0037] Optionally, in this embodiment of the application, the core capability characteristics of the supplier are concatenated with the corresponding dynamic performance characteristics to obtain a comprehensive evaluation feature vector; The comprehensive evaluation feature vector is input into a pre-trained supplier comprehensive evaluation model to obtain an initial evaluation score; The initial evaluation score is divided into levels based on a preset rating threshold to obtain the supplier's real-time rating result.

[0038] It should be noted that the training process of the aforementioned supplier comprehensive evaluation model includes: Obtain the core capability characteristics of suppliers and the corresponding dynamic performance characteristics of the period from several sets of historical cooperation data collected over historical periods; and use cooperation satisfaction as the tag value; The core capability characteristics of suppliers, their corresponding dynamic performance characteristics, and tag values ​​from several sets of historical cooperation data collected over several periods are grouped and labeled, denoted as follows: It is a natural number; Will The core capabilities of suppliers, their dynamic performance characteristics for the corresponding period, and tag values ​​from historical cooperation data collected over a historical period are used as sample data. Less than The natural numbers, and using the sample data, the mean of the sample data is obtained, denoted as the sample set; The core capability characteristics of suppliers, the dynamic performance characteristics of the corresponding period, and the tag values ​​in the historical cooperation data of the remaining historical collection periods are used as the test set; a training sample set is formed based on the sample set and the test set. The GBDT model and the neural network model are trained based on the training sample set, and the trained model is denoted as the supplier comprehensive evaluation model. The GBDT model is configured with the following basic parameters: the number of decision trees is set to 200, with the optimal value determined through grid search, and a test range of 100-500; the learning rate is set to 0.05 (to balance model fit based on the number of decision trees); the maximum depth is set to 10 to avoid overfitting, with a test range of 5-15; the minimum number of sample splits is set to 20; the minimum number of leaf node samples is set to 10; and the loss function is mean squared error, suitable for regression tasks. The training process involves initializing a single decision tree, fitting it to the training sample set data, and calculating the initial residual; adding decision trees one by one, fitting the residual of the previous tree to each tree, and updating the model parameters through gradient descent; evaluating the model performance on the validation set every 50 trees trained, and stopping training if there is no performance improvement for 10 consecutive rounds.

[0039] The basic parameters of the neural network model are set as follows: 128 neurons are set for the first hidden layer of the neural network model, the activation function is set to ReLU, and the dropout rate is set to 0.2 to prevent overfitting; 64 neurons are set for the second hidden layer, the activation function is set to ReLU, and the dropout rate is set to 0.2; 32 neurons are set for the third hidden layer, the activation function is set to ReLU; 1 neuron is set for the output layer of the neural network model, using a linear activation function; the optimizer is the Adam optimizer, the learning rate is set to 0.001, the first-order momentum decay coefficient is set to 0.9, and the second-order momentum decay coefficient is set to 0.999; the loss function is the mean squared error. During training, the batch size was set to 32 and the number of training epochs to 100. Training sample data was input into the input layer of the neural network model in batches, and forward propagation was used to calculate the predicted values. The MSE loss between the predicted values ​​and the true labels was calculated, and the weights were updated via backpropagation. After each training epoch, the model was evaluated using a validation set, and the model weights with the lowest validation set loss were saved. If the validation set loss increased for 15 consecutive epochs, an early stopping mechanism was triggered to prevent overfitting, and L2 regularization was applied to further suppress overfitting. The outputs of the two models were weighted and fused in a 6:4 ratio to obtain the final evaluation score.

[0040] To further explain, the preset rating thresholds are divided into 5 levels, including: Level A (e.g., 90-100 points, excellent supplier), Level B (e.g., 80-89 points, good supplier), Level C (e.g., 70-79 points, qualified supplier), Level D (e.g., 60-69 points, supplier with room for improvement), and Level E (e.g., below 60 points, unqualified supplier).

[0041] It should be further explained that, in the specific implementation process, the process of adjusting the weight of the real-time supplier rating results based on the priority of procurement needs to generate the supplier tier list includes: Optionally, in this application embodiment, current procurement demand information is obtained, and procurement demand priority features are extracted. The procurement demand priority features include, but are not limited to, demand urgency, material importance, quality requirement level, and cost sensitivity. The weight coefficients of each procurement demand priority feature are calculated based on the analytic hierarchy process. These weight coefficients include, but are not limited to, the urgency weight coefficient, the material importance weight coefficient, the quality requirement level weight coefficient, and the cost sensitivity weight coefficient. For example, the urgency weight of urgent procurement in the current procurement demand information is 0.4, and the material importance weight of key materials in the current procurement demand information is 0.35. The initial evaluation score of the supplier's real-time rating result is adjusted according to the weighting coefficients to obtain the adjusted evaluation score; It should be noted that the formula for the adjusted evaluation score is: Adjusted Evaluation Score = Initial Evaluation Score × (Requirement Urgency Weighting Coefficient × Urgency Fit Coefficient + Material Importance Weighting Coefficient × Importance Fit Coefficient + Quality Requirement Level Weighting Coefficient × Quality Fit Coefficient + Cost Sensitivity Weighting Coefficient × Cost Fit Coefficient); where the urgency fit coefficient, importance fit coefficient, quality fit coefficient, and cost fit coefficient are determined based on the degree of matching between supplier characteristics and procurement needs. If the matching degree is ≥80%, the corresponding fit coefficient is set to 1.2; if the matching degree is [60%, 80%), the corresponding fit coefficient is set to 1.0; if the matching degree is <60%, the corresponding fit coefficient is set to 1.0. The matching coefficient is set to 0.8. For example, for critical materials requiring urgent procurement, the urgency weight is 0.4, the material importance weight is 0.35, the quality requirement level weight coefficient is 0.15, and the cost sensitivity weight coefficient is 0.1. A certain A-level supplier's initial evaluation score is 92 points, and its urgency matching coefficient is 1.2, its importance matching coefficient is 1.2, its quality matching coefficient is 1.0, and its cost matching coefficient is 0.8. Then the adjusted evaluation score = 92 × (0.4 × 1.2 + 0.35 × 1.2 + 0.15 × 1.0 + 0.1 × 0.8) = 92 × 1.13 = 103.96 points. It is capped at 100 points and ranks higher than other suppliers in the tiered list.

[0042] The supplier classification is re-established based on the adjusted evaluation scores, and the suppliers are sorted by level and adjusted scores to generate a supplier classification list. The supplier classification list includes, but is not limited to, supplier name, rating level, adjusted evaluation score, core advantages, and type of needs that are suitable for the supplier.

[0043] It should be further explained that, in the specific implementation process, the process of constructing a supplier digital profile system based on the aforementioned supplier tiered list includes: It should be noted that the supplier digital profile system includes a core capability profile, a dynamic performance profile, and a cooperation adaptation profile. The core capability profile includes static core features such as basic qualifications, performance capabilities, quality control, service support, and cost adaptation. The dynamic performance profile includes dynamic features such as real-time order execution, quality status, and logistics progress. The cooperation adaptation profile includes scenario-based features such as adaptable demand types, optimal cooperation models, and supply capacity boundaries.

[0044] Optionally, in this embodiment of the application, the supplier name, rating level, adjusted evaluation score, core advantages and matching demand type of each supplier are extracted from the supplier classification list to construct a core capability profile; the dynamic performance characteristics of the corresponding supplier are obtained to construct a dynamic performance profile; and the semantic features in the cooperation evaluation text or complaint record are extracted based on natural language processing technology (NLP) to construct a cooperation matching profile. By integrating the core capability profile, dynamic performance profile, and cooperation adaptation profile of the corresponding supplier, a complete digital profile of the corresponding supplier is formed.

[0045] It should be further explained that, in the specific implementation process, the process of using density clustering algorithm to perform cooperative suitability clustering on suppliers in the profiling system to obtain the supplier classification cooperation model includes: Optionally, in this embodiment of the application, a cooperation adaptation feature set is extracted from the supplier digital profile system. The cooperation adaptation feature set includes, but is limited to, demand type adaptation, batch supply capability, customization capability, delivery cycle range, quality level adaptation, and cost range. Based on Z-score standardization, the cooperative adaptation feature set is standardized to obtain a standardized adaptation feature set. Cluster analysis is performed on the standardized fitting feature set based on the density clustering algorithm (DBSCAN) to obtain the corresponding clusters. The neighborhood radius in the density clustering algorithm is set to 0.8 and the minimum number of samples is set to 5. Feature analysis is performed on each cluster to extract the core features of the cluster. The core features of the cluster include, but are not limited to, large-batch standardized supply clusters, small-batch customized supply clusters, and emergency order response clusters. By assigning cooperation scenario tags and matching procurement demand types to each cluster, a supplier classification cooperation model is obtained.

[0046] It should be noted that the density clustering algorithm calculates the Euclidean distance between the fitting feature vectors of each supplier, and divides suppliers whose Euclidean distance is less than the neighborhood radius and whose number of samples in the neighborhood is greater than the minimum number of samples into the same cluster. Through the analysis of the core features of the clusters, the advantageous scenarios of suppliers in each category are identified. For example, the supplier characteristics of the emergency order response cluster are short delivery cycle, fast response speed, and high fault tolerance, which are suitable for procurement scenarios such as emergency replenishment and sudden demand.

[0047] It should be further explained that, in the specific implementation process, the process of intelligently matching procurement needs and optimizing cooperation strategies based on the aforementioned supplier classification cooperation model to generate customized cooperation solutions includes: It should be noted that existing supplier cooperation strategies are mostly fixed templates, lacking specificity and failing to consider the impact of dynamic market changes on cooperation, resulting in insufficient flexibility in cooperation plans and an inability to adapt to complex market environments. This application aims to achieve precise and adaptive management of supplier cooperation.

[0048] Optionally, in this embodiment of the application, detailed parameters of the current procurement needs are obtained, including but not limited to demand type, procurement quantity, delivery cycle requirements, quality standards, budget range, and preferred cooperation mode. Calculate the matching degree between the detailed parameters of the current procurement needs and the core features of each cluster in the supplier classification cooperation model, and select the target cluster with the highest matching degree. The top three suppliers with the highest adjusted evaluation scores from the target cluster are selected as candidate suppliers. For each candidate supplier, a customized cooperation strategy is formulated based on the core strengths and weaknesses in the corresponding supplier digital profile. The customized cooperation strategy includes, but is not limited to, procurement volume strategy, delivery cycle strategy, quality control strategy, cost optimization strategy, and risk prevention and control strategy. Based on the customized cooperation strategy, a corresponding customized cooperation plan is generated, and the customized cooperation plan is dynamically iterated and updated in combination with real-time market dynamics to achieve comprehensive management of the supplier's entire lifecycle.

[0049] Example 2: like Figure 2 As shown, a supplier integrated management system based on real-time AI analysis is disclosed. The system includes, but is not limited to, a feature extraction module, a real-time acquisition module, a comprehensive evaluation module, a digital profiling module, and a dynamic cooperation module. The feature extraction module is used to standardize the pre-collected historical cooperation data of suppliers and extract the core capability features of suppliers from the standardized historical cooperation data based on the deep factor decomposition model. The real-time acquisition module collects and verifies multi-dimensional data throughout the supplier cooperation process based on a multi-source real-time data acquisition terminal, obtains real-time operational data of suppliers, and extracts dynamic performance characteristics from the real-time operational data of suppliers through a time-series attention mechanism. The comprehensive evaluation module performs an AI-based comprehensive evaluation of dynamic performance characteristics based on the core capability characteristics of the suppliers, obtains real-time supplier rating results, and adjusts the weight of the real-time supplier rating results in combination with the priority of procurement needs to generate a supplier tier list. The digital profiling module constructs a supplier digital profiling system based on the supplier hierarchical directory, and uses a density clustering algorithm to perform cooperation-adaptive clustering on the suppliers in the profiling system to obtain a supplier classification cooperation model. The dynamic cooperation module intelligently matches procurement needs and optimizes cooperation strategies based on the supplier classification cooperation model, generates customized cooperation solutions, and dynamically iterates and updates the cooperation solutions in conjunction with real-time market dynamics, thereby realizing comprehensive management of the entire supplier lifecycle.

[0050] Optionally, in this embodiment, those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0051] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0052] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more electronic devices to execute all or part of the steps of the methods described in the various embodiments of this application.

[0053] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0054] In the several embodiments provided in this application, it should be understood that the disclosed application can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0055] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0056] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0057] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A comprehensive supplier management method based on real-time AI analysis, characterized in that, The method includes: The pre-collected historical cooperation data of suppliers is standardized, and the core capability characteristics of suppliers are extracted from the standardized historical cooperation data based on the deep factor decomposition model. Based on the multi-source real-time data acquisition terminal, multi-dimensional data collection and real-time verification are carried out on the entire process of supplier cooperation to obtain real-time operational data of suppliers, and dynamic performance characteristics are extracted from the real-time operational data of suppliers through the time-series attention mechanism. Based on the core capabilities of the suppliers, the dynamic performance characteristics are comprehensively evaluated by AI to obtain the real-time rating results of the suppliers. The weight of the real-time rating results of the suppliers is adjusted in combination with the priority of procurement needs to generate a supplier classification list. Based on the supplier classification directory, a supplier digital profile system is constructed, and the suppliers in the profile system are subjected to cooperation adaptability clustering using a density clustering algorithm to obtain a supplier classification cooperation model. Based on the supplier classification and cooperation model, procurement needs are intelligently matched and cooperation strategies are optimized to generate customized cooperation solutions. These customized cooperation solutions are then dynamically iterated and updated in conjunction with real-time market dynamics, thereby achieving comprehensive management of the entire supplier lifecycle.

2. The supplier comprehensive management method based on real-time AI analysis according to claim 1, characterized in that, The process of extracting core capability features of suppliers from standardized historical supplier cooperation data based on a deep factorization model includes: Multi-dimensional historical feature data is extracted from standardized supplier historical cooperation data, and the multi-dimensional historical feature data is discretized and normalized based on a preset feature dimension system to obtain the supplier basic feature set. First-order feature filtering is performed on the supplier basic feature set to obtain an effective basic feature set; cross feature data is extracted from the multi-dimensional historical feature data, and feature embedding is performed on the cross feature data to obtain the supplier cross feature set; The effective basic feature set and the supplier cross feature set are input into the deep factorization model to obtain the first-order feature interaction vector and the higher-order feature interaction vector. The attention weight distribution of the first-order feature interaction vector and the higher-order feature interaction vector is calculated based on the feature interaction attention layer to obtain the feature interaction weight. The first-order feature interaction vector and the higher-order feature interaction vector are weighted and fused based on the feature interaction weights to obtain a fused feature vector; the fused feature vector is then subjected to feature dimensionality reduction to obtain the supplier's core capability features.

3. The supplier integrated management method based on real-time AI analysis according to claim 2, characterized in that, The process of obtaining real-time operational data from suppliers by collecting and verifying multi-dimensional data across the entire supplier cooperation process using multi-source real-time data acquisition terminals includes: The system obtains real-time order data by acquiring supplier order execution data through the order management system interface and adding timestamps; it also obtains real-time quality data by acquiring process quality data and batch sampling data from the product manufacturing process through the quality inspection equipment data acquisition interface and adding timestamps; it obtains real-time logistics data by acquiring cargo transportation data through the logistics GPS tracking interface and adding timestamps; it obtains real-time service data by acquiring service interaction data through the after-sales service platform interface and adding timestamps; and it obtains real-time financial data by acquiring financial interaction data through the financial settlement system interface and adding timestamps. The real-time order data, real-time quality data, real-time logistics data, real-time service data, and real-time financial data are aligned with time windows and correlated to obtain real-time multi-source fused data; based on preset verification rules, the real-time multi-source fused data is cross-validated to obtain real-time supplier operation data.

4. The supplier integrated management method based on real-time AI analysis according to claim 3, characterized in that, The process of extracting dynamic fulfillment features from supplier real-time operational data using a temporal attention mechanism includes: Based on the temporal attention mechanism, real-time dynamic data is extracted from the supplier's real-time operation data, and the real-time dynamic data is sorted in chronological order to obtain a real-time temporal data sequence; local temporal features are extracted from the real-time temporal data sequence using a temporal attention convolutional layer to obtain local temporal features. The attention weights of each time step in the local temporal features are calculated based on the temporal attention encoding layer to obtain weighted temporal features; global average pooling is then performed on the weighted temporal features to obtain dynamic performance features.

5. The supplier integrated management method based on real-time AI analysis according to claim 4, characterized in that, The process of obtaining a real-time supplier rating result by performing a comprehensive AI evaluation of dynamic performance characteristics based on the supplier's core competency characteristics includes: The supplier's core capability features are concatenated with the corresponding dynamic performance features to obtain a comprehensive evaluation feature vector; the comprehensive evaluation feature vector is input into a pre-trained supplier comprehensive evaluation model to obtain an initial evaluation score; the initial evaluation score is classified into levels based on a preset rating threshold to obtain the supplier's real-time rating result.

6. The supplier integrated management method based on real-time AI analysis according to claim 5, characterized in that, The training process of the supplier comprehensive evaluation model includes: Obtain the core capability characteristics of suppliers and the corresponding dynamic performance characteristics of the period from several sets of historical cooperation data collected over historical periods; and use cooperation satisfaction as the tag value; The core capability characteristics of suppliers, their corresponding dynamic performance characteristics, and tag values ​​from several sets of historical cooperation data collected over several periods are grouped and labeled, denoted as follows: It is a natural number; Will The core capabilities of suppliers, their dynamic performance characteristics for the corresponding period, and tag values ​​from historical cooperation data collected over a historical period are used as sample data. Less than The natural numbers, and using the sample data, the mean of the sample data is obtained, denoted as the sample set; The core capability characteristics of suppliers, the dynamic performance characteristics of the corresponding period, and the tag values ​​in the historical cooperation data of the remaining historical collection periods are used as the test set; a training sample set is formed based on the sample set and the test set. The GBDT model and the neural network model are trained based on the training sample set, and the trained model is denoted as the supplier comprehensive evaluation model.

7. The supplier integrated management method based on real-time AI analysis according to claim 6, characterized in that, The process of adjusting the weights of real-time supplier ratings based on procurement needs to generate a tiered supplier list includes: Obtain current procurement demand information and extract procurement demand priority features; calculate the weight coefficients of each procurement demand priority feature based on the analytic hierarchy process (AHP). The initial evaluation score of the supplier's real-time rating result is adjusted according to the weighting coefficients to obtain the adjusted evaluation score; The supplier grading list is generated by reclassifying the suppliers based on the adjusted evaluation scores and sorting them according to the grade order and the adjusted scores.

8. The supplier integrated management method based on real-time AI analysis according to claim 7, characterized in that, The process of constructing a supplier digital profile system based on the aforementioned supplier tiered directory includes: Extract the supplier name, rating level, adjusted evaluation score, core advantages, and suitable demand type of each supplier from the supplier classification list to construct a core capability profile; obtain the dynamic performance characteristics of the corresponding supplier to construct a dynamic performance profile; and extract semantic features from cooperation evaluation texts or complaint records based on natural language processing technology to construct a cooperation suitability profile. By integrating the core capability profile, dynamic performance profile, and cooperation adaptation profile of the corresponding supplier, a complete digital profile of the corresponding supplier is formed.

9. A supplier integrated management method based on real-time AI analysis according to claim 8, characterized in that, The process of obtaining a supplier classification cooperation model by performing cooperation-adaptive clustering of suppliers in the profiling system using density clustering algorithms includes: Extract the cooperation adaptation feature set from the supplier digital profile system; based on Z-score standardization, perform feature standardization on the cooperation adaptation feature set to obtain the standardized adaptation feature set; Cluster analysis is performed on the standardized adaptation feature set based on the density clustering algorithm to obtain the corresponding clusters; feature analysis is performed on each cluster to extract the core features of the cluster, which include large-batch standardized supply clusters, small-batch customized supply clusters, and emergency order response clusters; By assigning cooperation scenario tags and matching procurement demand types to each cluster, a supplier classification cooperation model is obtained.

10. A supplier integrated management system based on real-time AI analysis, implementing the supplier integrated management method based on real-time AI analysis as described in any one of claims 1 to 9, characterized in that, include: The module includes a feature extraction module, a real-time acquisition module, a comprehensive evaluation module, a digital profiling module, and a dynamic collaboration module. The feature extraction module is used to standardize the pre-collected historical cooperation data of suppliers and extract the core capability features of suppliers from the standardized historical cooperation data based on the deep factor decomposition model. The real-time acquisition module collects and verifies multi-dimensional data throughout the entire supplier cooperation process based on a multi-source real-time data acquisition terminal, obtains real-time operational data of suppliers, and extracts dynamic performance characteristics from the real-time operational data of suppliers through a time-series attention mechanism. The comprehensive evaluation module performs an AI-based comprehensive evaluation of dynamic performance characteristics based on the core capability characteristics of the suppliers, obtains real-time supplier rating results, and adjusts the weight of the real-time supplier rating results in combination with the priority of procurement needs to generate a supplier tier list. The digital profiling module constructs a supplier digital profiling system based on the supplier hierarchical directory, and uses a density clustering algorithm to perform cooperation-adaptive clustering on the suppliers in the profiling system to obtain a supplier classification cooperation model. The dynamic cooperation module intelligently matches procurement needs and optimizes cooperation strategies based on the supplier classification cooperation model, generates customized cooperation solutions, and dynamically iterates and updates the cooperation solutions in conjunction with real-time market dynamics, thereby realizing comprehensive management of the entire supplier lifecycle.

Images