Supplier intelligent portrait construction method, system and equipment based on deep learning

Abstract

The invention discloses a supplier intelligent portrait construction method, system and device based on deep learning, and the method comprises the steps: collecting data from multiple dimensions through a distributed crawler technology, and carrying out the cleaning, fusion and time sequence analysis of the data; carrying out credibility evaluation on the collected data, and marking a credibility label on each data point; supplier features are extracted from multiple dimensions, screening is performed from each dimension through a feature selection algorithm, typical feature subsets are formed, nonlinear combination and weight distribution are performed on the typical feature subsets through a deep learning model, and multi-dimensional feature vectors of suppliers are output and obtained; the multi-dimensional feature vectors comprise multi-dimensional feature vectors in a product management aspect and a green environmental protection information aspect, and a supplier full-dimensional portrait is constructed according to the multi-dimensional feature vectors in the product management aspect; and performing green tagging processing according to the multi-dimensional feature vector in the aspect of green environmental protection information to obtain a supplier full-dimensional portrait with a green tag.

Description

Technical Field

[0001] The present invention relates to the field of supplier portrait technology, and in particular to a method, system and device for constructing supplier intelligent portrait based on deep learning. Background Art

[0002] Currently, with the changes in the market environment, the risks and uncertainties of the supply chain are increasing. There are great uncertainties in the procurement, production and logistics cycles of suppliers, and there are uncertainties in product management and cycles. Especially for small and medium-sized enterprises with weak supplier control capabilities, this uncertainty brings great management and delivery difficulties. For large enterprises, they can use their own advantageous position in the supply chain to require suppliers to strictly implement procurement plans. For the majority of small and medium-sized enterprises, on the one hand, the procurement volume is small, the procurement volume is not stable enough, and the ability to control suppliers is weak; on the other hand, there is not enough money to build a complete supplier management system to obtain real-time information on suppliers' procurement, production, quality inspection, etc.

[0003] Therefore, small and medium-sized enterprises rely more on the experience of buyers to implement supply chain management, which lacks objectivity and makes it difficult to comprehensively and objectively evaluate the comprehensive capabilities of suppliers, which can easily lead to management and delivery difficulties. Summary of the invention

[0004] The present invention provides a method for constructing an intelligent supplier portrait based on deep learning to solve the technical problems that it is difficult to comprehensively and objectively evaluate the comprehensive capabilities of suppliers and that easily lead to management and delivery difficulties.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a method for constructing a supplier intelligent portrait based on deep learning, and the method includes:

[0006] Acquire an initial data set from multiple data sources, and clean, merge, and perform time series analysis on the initial data set to obtain a supplier's full life cycle quality data set, wherein the initial data set includes data on product management and green environmental protection information;

[0007] Conducting a credibility assessment on the supplier's full life cycle quality data set, and labeling each data point with a credibility label to obtain a supplier quality data set with a credibility label;

[0008] Extract supplier features from multiple dimensions based on the supplier quality data set with credibility labels, screen out typical features in each dimension through a feature selection algorithm and generate a typical feature subset, perform nonlinear combination and weight assignment on the typical feature subset through a deep learning model, and output a multi-dimensional feature vector of the supplier, the multi-dimensional feature vector includes a multi-dimensional feature vector in product management and a multi-dimensional feature vector in green environmental protection information, and construct a full-dimensional portrait of the supplier based on the multi-dimensional feature vector in product management;

[0009] Green labeling is performed based on the multi-dimensional feature vector of the green environmental protection information to obtain a full-dimensional portrait of the supplier with a green label.

[0010] The specific steps of obtaining the initial data set from multiple data sources, and performing cleaning, fusion and time series analysis on the initial data set to obtain the supplier's full life cycle quality data set include:

[0011] Further, data on the product management and green environmental protection information are collected from the plurality of data sources, and all the collected data are pre-processed by a data cleaning algorithm to remove duplicate, erroneous and invalid data, thereby obtaining a supplier quality data set after preliminary cleaning;

[0012] The key features of the supplier quality data set are extracted by feature engineering methods. The same data from different sources are fused using a multi-source data fusion algorithm based on the key features. The data format and unit are unified through data standardization. Missing values ​​are filled using data completion technology to obtain a consistent and complete fused supplier full-dimensional data set.

[0013] The time series analysis method is used to identify the time characteristics and change trends of the supplier's full-dimensional data set, and the interpolation algorithm is used to repair data missing and inconsistency problems, so as to obtain a supplier's full life cycle quality data set with temporal integrity.

[0014] Furthermore, the specific steps of performing credibility assessment on the supplier's full life cycle quality data set and labeling each data point with a credibility label to obtain the supplier quality data set with a credibility label include:

[0015] Performing consistency scoring on the supplier's full life cycle quality data set based on a data consistency scoring model, and rechecking the data source if the consistency score is lower than a threshold;

[0016] Identify potential abnormal data in the supplier's full life cycle quality data set through an anomaly detection algorithm, and filter out the abnormal data;

[0017] Using a data accuracy scoring algorithm to score the accuracy of the supplier's full life cycle quality data set after screening, to obtain an accuracy index for each piece of data;

[0018] Use the data integrity scoring model to score the integrity of the supplier's full life cycle quality data set after screening, determine whether the information in the supplier's full life cycle quality data set after screening is comprehensive, and verify the original data source again for the incomplete data found to supplement and improve the information;

[0019] Comprehensive consistency score, accuracy score and completeness score, and use weighted average method to give a comprehensive credibility score to each data in the supplier's full life cycle quality data set after screening;

[0020] Determine whether the comprehensive credibility score of each data in the supplier's full life cycle quality data set exceeds a preset threshold. If so, add a credibility label to each data exceeding the preset threshold to obtain a supplier quality data set with a credibility label.

[0021] Furthermore, the supplier characteristics are extracted from multiple dimensions based on the supplier quality data set with credibility labels, and typical characteristics of each dimension are screened out through a feature selection algorithm to generate a typical feature subset, and the typical feature subset is nonlinearly combined and weighted by a deep learning model to output a multidimensional feature vector of the supplier, wherein the multidimensional feature vector includes a multidimensional feature vector of product management and a multidimensional feature vector of green environmental protection information. The specific steps of constructing a full-dimensional portrait of the supplier according to the multidimensional feature vector of product management include:

[0022] The supplier quality data set with credibility labels includes supplier quality data on product management and green environmental protection information, and multi-dimensional supplier features are extracted using multi-dimensional feature extraction technology, wherein the supplier quality data on product management includes quality, delivery and cost dimensions, and the supplier quality data on green environmental protection information includes green procurement, green manufacturing, green management, energy-saving and emission reduction measures and waste treatment dimensions;

[0023] Use principal component analysis or random forest algorithm to screen out typical features in each dimension and generate a typical feature subset;

[0024] A convolutional neural network is used to perform nonlinear combination and weight allocation on the selected typical feature subsets, and a historical data set is used to train the model. By adjusting the network structure and parameters, a multi-dimensional feature vector of the supplier is output. The multi-dimensional feature vector of product management includes quality, delivery and cost dimension feature vectors, and the multi-dimensional feature vector of green environmental information includes green procurement, green manufacturing, green management, energy-saving and emission reduction measures, and waste treatment dimension feature vectors;

[0025] A full-dimensional portrait of the supplier is constructed based on the multi-dimensional feature vectors of the product management.

[0026] Furthermore, the specific steps of performing green labeling processing according to the multi-dimensional feature vector of the green environmental protection information to obtain a full-dimensional portrait of the supplier with a green label include:

[0027] According to the multi-dimensional feature vector of the green environmental protection information, the suppliers are divided into different green levels by using a clustering algorithm;

[0028] Use fuzzy reasoning technology to rate suppliers’ greenness;

[0029] Integrate the green grade and the green degree score into a green label, and label the supplier with the green label;

[0030] A full-dimensional portrait of the supplier with a green label is formed based on the multi-dimensional feature vector of the green environmental protection information and the green label, and the performance and score of each supplier in each dimension are recorded and displayed in detail.

[0031] Furthermore, the green grading of suppliers also includes:

[0032] Use the decision tree model to build a classification model, and use the supplier data marked with green labels to train the classification model;

[0033] For new suppliers, the multi-dimensional feature vector of the new supplier's green environmental information is output by the trained classification model to obtain a predicted green grade.

[0034] The embodiment of the present invention also provides a supplier intelligent portrait construction system based on deep learning, the system comprising:

[0035] A processing module is used to obtain an initial data set from multiple data sources, and clean, merge and perform time series analysis on the initial data set to obtain a supplier's full life cycle quality data set, wherein the initial data set includes data on product management and green environmental protection information;

[0036] A credibility evaluation module is used to perform credibility evaluation on the supplier's full life cycle quality data set and label each data point with a credibility label to obtain a supplier quality data set with a credibility label;

[0037] A product management portrait construction module is used to extract supplier features from multiple dimensions based on the supplier quality data set with credibility labels, and screen out typical features in each dimension through a feature selection algorithm and generate a typical feature subset, and perform nonlinear combination and weight assignment on the typical feature subset through a deep learning model to output a multi-dimensional feature vector of the supplier, wherein the multi-dimensional feature vector includes a multi-dimensional feature vector of product management and a multi-dimensional feature vector of green environmental protection information, and a full-dimensional portrait of the supplier is constructed based on the multi-dimensional feature vector of product management;

[0038] The green environmental protection portrait construction module is used to perform green labeling processing based on the multi-dimensional feature vector of the green environmental protection information to obtain a full-dimensional portrait of the supplier with a green label.

[0039] Furthermore, the processing module includes:

[0040] A data cleaning submodule is used to collect data on the product management and green environmental protection information from multiple data sources, and pre-process all collected data through a data cleaning algorithm to remove duplicate, erroneous and invalid data to obtain a supplier quality data set after preliminary cleaning;

[0041] The data fusion submodule is used to extract the key features of the supplier quality data set by using a feature engineering method, and to fuse the same data from different sources using a multi-source data fusion algorithm according to the key features, unify the data format and unit through data standardization, and fill in the missing values ​​using data completion technology to obtain a fused supplier full-dimensional data set with consistency and integrity;

[0042] The time series analysis submodule is used to identify the time characteristics and change trends of the supplier's full-dimensional data set using a time series analysis method, and use an interpolation algorithm to repair data missing and inconsistency problems to obtain a supplier's full life cycle quality data set with time series integrity.

[0043] Furthermore, the credibility assessment module includes:

[0044] A consistency scoring submodule, used to score the consistency of the supplier's full life cycle quality data set based on a data consistency scoring model, and if the consistency score is lower than a threshold, recheck the data source;

[0045] An abnormal data identification submodule is used to identify potential abnormal data in the supplier's full life cycle quality data set through an abnormality detection algorithm and filter out the abnormal data;

[0046] The accuracy scoring submodule is used to use the data accuracy scoring algorithm to perform accuracy scoring on the screened supplier's full life cycle quality data set to obtain the accuracy index of each piece of data;

[0047] The integrity scoring submodule is used to use the data integrity scoring model to perform integrity scoring on the supplier's full life cycle quality data set after screening, to determine whether the information in the supplier's full life cycle quality data set after screening is comprehensive, and to verify the original data source again for the incomplete data found to supplement and improve the information;

[0048] The comprehensive scoring submodule is used to comprehensively score the consistency, accuracy and completeness, and use the weighted average method to perform a comprehensive credibility score on each data in the supplier's full life cycle quality data set after screening;

[0049] The credibility label submodule is used to determine whether the comprehensive credibility score of each data in the supplier's full life cycle quality data set exceeds a preset threshold. If so, a credibility label is added to each data exceeding the preset threshold to obtain a supplier quality data set with a credibility label.

[0050] Furthermore, the product management portrait construction module includes:

[0051] A feature extraction submodule is used for the supplier quality data set with credibility labels, including supplier quality data on product management and green environmental protection information, and adopts multi-dimensional feature extraction technology to extract multi-dimensional supplier features, wherein the supplier quality data on product management includes quality, delivery and cost dimensions, and the supplier quality data on green environmental protection information includes green procurement, green manufacturing, green management, energy-saving and emission reduction measures and waste treatment dimensions;

[0052] The feature screening submodule is used to screen out typical features in each dimension and generate a typical feature subset using principal component analysis or random forest algorithm;

[0053] A multi-dimensional feature vector output submodule, which is used to use a convolutional neural network to perform nonlinear combination and weight allocation on the selected typical feature subsets, use historical data sets to train the model, and output the multi-dimensional feature vector of the supplier by adjusting the network structure and parameters. The multi-dimensional feature vector of product management includes quality, delivery and cost dimension feature vectors, and the multi-dimensional feature vector of green environmental information includes green procurement, green manufacturing, green management, energy-saving and emission reduction measures, and waste treatment dimension feature vectors;

[0054] The first construction submodule is used to construct a full-dimensional portrait of the supplier based on the multi-dimensional feature vector of the product management aspect.

[0055] Furthermore, the green environmental protection portrait construction module includes:

[0056] A clustering submodule, for classifying suppliers into different green levels using a clustering algorithm according to the multi-dimensional feature vector of the green environmental protection information;

[0057] Output green degree scoring submodule, used to score the green degree of suppliers using fuzzy reasoning technology;

[0058] A green label submodule, for integrating the green grade and the green degree score into a green label, and labeling the supplier with the green label;

[0059] The second construction submodule is used to form a full-dimensional portrait of the supplier with a green label based on the multi-dimensional feature vector of the green environmental protection information and the green label, and record and display the performance and score of each supplier in each dimension in detail.

[0060] An embodiment of the present invention further provides a computer device, including:

[0061] one or more processors;

[0062] A storage device for storing one or more programs;

[0063] When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement any of the above methods.

[0064] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program, wherein the program implements any of the above-mentioned methods when executed by a processor.

[0065] An embodiment of the present invention further provides a computer program product, including computer program instructions, wherein the computer program instructions enable a computer to execute any one of the above methods.

[0066] Beneficial effects of the embodiments of the present invention:

[0067] By collecting data from multiple dimensions through distributed crawler technology and cleaning, fusion and time series analysis of the data, the validity of the data can be ensured. By evaluating the credibility of the data, the reliability and quality of the data can be ensured. By using deep learning models, comprehensive evaluations of suppliers are conducted from multiple dimensions, and a full-dimensional portrait of the supplier is constructed. This can provide a more comprehensive understanding of the comprehensive capabilities of the supplier, help enterprises select suitable suppliers more objectively, and ensure the supplier's management and delivery capabilities.

[0068] At the same time, the present invention also performs green labeling on the multi-dimensional feature vectors of the supplier's green environmental protection information, constructs a full-dimensional portrait of the supplier with a green label, so as to classify the suppliers into green suppliers and non-green suppliers, and displays the performance and score of each supplier in various dimensions by integrating the green grades and green degree scores of the suppliers after green labeling, so as to provide a basis for the green procurement of enterprises. In addition, the present invention uniformly manages the full-dimensional portrait of suppliers and the full-dimensional portrait of suppliers with green labels, which can promote the refinement and greening of supply chain management, further help enterprises optimize supplier resources, enable the present invention to more comprehensively and objectively display the types and performance of suppliers, improve the efficiency of supplier management, and promote the sustainable development strategy of enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 is a flow chart of an embodiment of the present invention;

[0070] Figure 2 is a flowchart of a specific implementation of step 10 of the present invention;

[0071] Figure 3 is a flowchart of a specific implementation of step 20 of the present invention;

[0072] Figure 4 is a flowchart of a specific implementation of step 30 of the present invention;

[0073] Figure 5 is a flowchart of a specific implementation of step 40 of the present invention;

[0074] Figure 6 is a schematic block diagram of an example computer device of an embodiment of the present invention.

[0075] The reference numerals of the drawings in the specification are as follows:

[0076] 600, computer device; 601, computing unit; 602, ROM; 603, RAM; 604, bus; 605, I / O interface; 606, input unit; 607, output unit; 608, storage unit; 609, communication unit. DETAILED DESCRIPTION

[0077] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0078] Embodiment 1

[0079] Please refer to Figures 1 to 5 , a method for constructing a supplier intelligent portrait based on deep learning is provided in an embodiment of the present invention, and the method includes:

[0080] Step S10: Acquire an initial data set from multiple data sources, and clean, fuse and perform time series analysis on the initial data set to obtain a supplier's full life cycle quality data set, wherein the initial data set includes data on product management and green environmental protection information.

[0081] Specifically, multiple data sources may include the supplier's official website, industry databases, and third-party evaluation reports, etc.

[0082] Distributed crawler technology is used to obtain initial data sets from multiple data sources, such as collecting product management data on supplier qualifications, product quality, delivery performance, and service quality, as well as collecting data on green environmental protection information such as green procurement, green manufacturing, green management, energy-saving and emission reduction measures, and waste disposal. Data cleaning tools such as Ta l end are used to remove duplicates, entity recognition and linking technology are used for fusion, and the K-nearest neighbor algorithm (K-NN) is used to fill missing values. Time series analysis is performed on these data, such as using Python's Pandas library to timestamp and sort the data to ensure the latest data, thereby obtaining a supplier's full life cycle quality data set with temporal integrity.

[0083] Step S10 specifically includes: Step S11: Collect data on product management and green environmental protection information from multiple data sources, and pre-process all collected data through data cleaning algorithms to remove duplicate, erroneous and invalid data to obtain a supplier quality data set after preliminary cleaning. In this way, the preliminary screening and purification of data quality is guaranteed, and the quality and availability of data are improved.

[0084] Specifically, the basic information of suppliers is first crawled through the distributed crawler technology crawling algorithm, including the supplier's name, address, and contact information, etc. Regular expressions are used to match the supplier information format, and its authenticity is confirmed through database comparison. If the data source contains complete basic information of the supplier, the next step is to crawl the supplier's product management data.

[0085] Data is indexed based on basic information such as supplier name, address and contact information, and distributed crawler technology is used to capture detailed product management data of each supplier in parallel, including basic product information, quality information, qualification information, evaluation information, delivery information, after-sales service information, cooperation information, innovation information, risk information, etc.

[0086] Crawl the basic product information of suppliers. Specifically, the basic product information includes product name, specification model and quantity, etc. Use natural language processing technology (NLP) to extract product specifications from web page data, and then use logical judgment algorithms to determine the completeness and accuracy of this information, and clarify the specification model and specific quantity. If the product information is complete and the specification model is clear, it is judged as valid data and enters the crawling of quality inspection results.

[0087] Crawl the product quality inspection report, identify the inspection items and inspection values ​​in the report through image recognition technology OCR (Optical Character Recognition), and automatically analyze whether the inspection items and inspection values ​​meet the preset standards according to the preset standards to determine whether the product quality is qualified. The judgment results are used for the next step of data analysis.

[0088] Capture the supplier's certification certificates and qualification documents, use data cleaning algorithms to filter out invalid data that is expired or unofficially certified, and ensure that the remaining qualification information is authentic and valid.

[0089] Collect supplier evaluation information such as supplier historical performance scores, user reviews and complaint records, use sentiment analysis technology to analyze user reviews and complaint records, extract sentiment tendencies and calculate the average score. For example, by analyzing the key words "satisfied" or "disappointed" in the comments to evaluate the sentiment color, and combine this with the score data (such as 5 / 5) to comprehensively judge the supplier's service level.

[0090] Distributed crawler technology is used to crawl detailed data on suppliers' green environmental information, including suppliers' environmental certification, green procurement, green manufacturing, green management, energy-saving and emission reduction measures, waste disposal and other green environmental related data. When collecting suppliers' green environmental data, automated scripts can be used to extract environmental certification information from documents provided by suppliers, such as ISO 14001 or other international certifications.

[0091] Green procurement means implementing green procurement standards during the procurement process, such as ISO 14001 environmental management system certification, green product certification, etc. Try to use environmentally friendly outer packaging materials or component materials, such as renewable resources, degradable materials and recyclable materials, to replace traditional materials and reduce the impact on the environment. Try to find substitutes for harmful materials and make full use of the company's excess equipment and materials.

[0092] Green manufacturing can be achieved by adopting optimized environmentally friendly production processes and advanced technologies to reduce energy and resource consumption and improve resource utilization. Waste treatment can establish a recycling system, conduct green recycling, achieve product recycling or other final disposal methods, such as landfill, incineration, etc.

[0093] Energy conservation and emission reduction measures are analyzed through natural language processing technology, such as the use of energy-saving equipment, optimization of production processes, and improvement of energy structure, to identify specific measures and implementation effects. For example, word frequency analysis is used to identify keywords such as "solar energy" and "wind energy" to quantify their energy-saving effects. If the data is found to deviate from the industry benchmark (such as an average annual energy saving rate of 5%), a further verification process is automatically triggered, and the API interface is called to query the relevant certification body to confirm the authenticity of the certificate.

[0094] Step S12: Use feature engineering methods to extract key features of the supplier quality data set, use multi-source data fusion algorithms to fuse the same data from different sources based on the key features, unify the data format and unit through data standardization, and use data completion technology to fill in missing values ​​to obtain a consistent and complete fused supplier full-dimensional data set.

[0095] Specifically, data fusion is performed on the cleaned supplier quality data sets, and these scattered supplier quality data sets are integrated through data fusion technology.

[0096] Use multi-source data fusion algorithms, such as entity recognition and linking technology, to match and integrate information from different data sources. Through entity recognition and linking technology, key information such as company name and address in various supplier data is identified, and the identified entities are compared with the existing database. If the information of the same company is found to be scattered in different data sets, similarity matching algorithms are used to merge them to ensure that each supplier only retains one record containing comprehensive information.

[0097] Through data standardization, the data format and unit are unified. The Z-score standardization method is used to adjust the inconsistency of data format and measurement unit.

[0098] Use data completion technology to fill in missing values ​​in the data set. If an important piece of data, such as net profit information, is missing, it can be supplemented through historical data trend forecasts or similar supplier data.

[0099] In a specific embodiment, if the net profit information of a supplier is not recorded in detail, the K-nearest neighbor (K-NN) algorithm is used to find the historical data of similar suppliers to predict the missing financial information.

[0100] Step S13: Use the time series analysis method to identify the time characteristics and change trends of the supplier's full-dimensional data set, and use the interpolation algorithm to repair data missing and inconsistency problems to obtain a supplier's full life cycle quality data set with temporal integrity.

[0101] Through the sliding window technology, the historical data of long-term cooperative suppliers are processed in segments to obtain data performance in different time windows. For example, each window is set to one quarter, the purchase amount and purchase frequency in each time window are calculated, and the ARI MA model (autoregressive integrated moving average model) is used to predict the intensity of economic activities in each time window, so as to analyze the economic change trend of the supplier.

[0102] If data is missing or inconsistent, the KNN interpolation algorithm can be used to repair the missing data to ensure the integrity and accuracy of the data. This step method can effectively predict its future cooperation potential and risks, thereby providing a scientific basis for corporate procurement decisions.

[0103] Step S20: Conduct credibility assessment on the supplier's full life cycle quality data set, and label each data point with a credibility label to obtain a supplier quality data set with a credibility label.

[0104] Specifically, a data quality evaluation model can be used to quantitatively analyze the accuracy, consistency, and completeness of the data.

[0105] Step S20 specifically includes: Step S21: Perform consistency scoring on the supplier's full life cycle quality data set based on the data consistency scoring model. If the consistency score is lower than the threshold, re-check the data source.

[0106] In a specific embodiment, the threshold may be set to 70%, that is, if the consistency score is lower than 70%, the data source is rechecked.

[0107] Step S22: Identify potential abnormal data in the supplier's full life cycle quality data set through an anomaly detection algorithm and filter out the abnormal data.

[0108] Specifically, using the Z-score-based method to identify abnormal points or abnormal patterns in the data helps to isolate possible data errors or external interference. For example, if a supplier's product return rate suddenly increases to 30%, which is much higher than its historical average of 5%, it is considered abnormal.

[0109] Step S23: Use a data accuracy scoring algorithm to score the accuracy of the screened supplier life cycle quality data set to obtain the accuracy index of each piece of data.

[0110] The data accuracy scoring algorithm may be a random forest algorithm, which is used to evaluate the accuracy of each piece of data after removing abnormal data. Specifically, data with an accuracy score higher than 80% is selected for subsequent analysis.

[0111] Step S24: Use the data integrity scoring model to perform integrity scoring on the screened supplier life cycle quality data set to determine whether the information in the screened supplier life cycle quality data set is comprehensive. For any incomplete data found, verify the original data source again to supplement and improve the information.

[0112] The data integrity scoring model can be a Gini coefficient calculation model. The Gini coefficient calculation model is used to evaluate the integrity of the data set. If the data is found to be incomplete, for example, the missing ratio of a supplier's qualification information exceeds 20%, it is necessary to return to the data source for supplementation.

[0113] Step S25: Comprehensive consistency score, accuracy score and completeness score, use weighted average method to make a comprehensive credibility score for each data in the screened supplier's full life cycle quality data set. The reliability of the data can be intuitively seen through this scoring system.

[0114] Preferably, data traceability technology can also be used to confirm whether the data with high credibility scores is authentic and available, such as using the SHA-256 algorithm to hash the data and compare it with the hash value of the original data source to verify the originality of the data. Blockchain is used to record each step of data capture and processing to ensure the reliability and correctness of the data, providing a solid foundation for establishing a high-quality supplier quality data set.

[0115] Step S26: Determine whether the comprehensive credibility score of each data in the supplier's full life cycle quality data set exceeds a preset threshold. If so, add a credibility label to each data that exceeds the preset threshold to obtain a supplier quality data set with a credibility label.

[0116] The preset threshold may be set to 70 points, and a comprehensive credibility score above 70 points may be judged as a high credibility score, and thus may be labeled as such.

[0117] Step S30: Extract supplier characteristics from multiple dimensions based on the supplier quality data set with credibility labels, and use the feature selection algorithm to screen out typical features in each dimension and generate a typical feature subset. Use the deep learning model to perform nonlinear combination and weight assignment on the typical feature subset, and output a multidimensional feature vector of the supplier. The multidimensional feature vector includes a multidimensional feature vector in product management and a multidimensional feature vector in green environmental information. A full-dimensional portrait of the supplier is constructed based on the multidimensional feature vector in product management.

[0118] The present invention constructs a full-dimensional portrait of suppliers and performs green labeling, conducts detailed multi-dimensional feature analysis and green level assessment of suppliers, thereby helping enterprises to more accurately identify and select suppliers that meet green and high-quality supply standards.

[0119] Step S30 specifically includes: Step S31: The supplier quality data set with a credibility label includes supplier quality data on product management and green environmental protection information, and multi-dimensional feature extraction technology is used to extract multi-dimensional supplier features, among which the supplier quality data on product management includes quality, delivery and cost dimensions, and the supplier quality data on green environmental protection information includes green procurement, green manufacturing, green management, energy-saving and emission reduction measures and waste treatment dimensions.

[0120] In the supplier quality dataset with credibility labels, multidimensional feature extraction technology is used to extract supplier characteristics in product management from the dimensions of quality, delivery and cost, and to extract supplier characteristics in green environmental protection information from the dimensions of green procurement, green manufacturing, green management, energy-saving and emission reduction measures and waste disposal.

[0121] Specifically, through statistical analysis methods, key indicators are extracted from the historical quality dimension, delivery dimension, cost dimension, service dimension, partnership dimension, green procurement dimension, green manufacturing dimension, green management dimension, energy-saving and emission reduction measures dimension, and waste disposal dimension.

[0122] The key indicators of historical quality dimension data include product qualification rate and quality accident frequency. The qualification rate is calculated by dividing the number of qualified products by the total number of products, and the quality accident frequency is calculated by counting the number of quality accidents within a certain period of time.

[0123] The key indicators of delivery dimension data include characteristics such as on-time order delivery rate and delivery cycle. The on-time order delivery rate is calculated by dividing the number of orders completed on time by the total number of orders. The delivery cycle requires calculating the average and standard deviation of the delivery cycle. The key indicators of the delivery dimension help evaluate the delivery efficiency and stability of suppliers.

[0124] Key indicators of cost dimension data include average purchase price and price volatility. The average purchase price is calculated by dividing the total amount of all purchase orders by the total quantity, and the price volatility is calculated as the ratio of the standard deviation of the price to the average. Key indicators of cost dimension help evaluate the price competitiveness of suppliers in the market.

[0125] The key indicators of service dimension data include after-sales service response time and customer satisfaction. After-sales service response time requires calculating the average and standard deviation of after-sales service response time, and customer satisfaction is calculated by evaluating the average of the survey scores. The key indicators of the service dimension help understand the supplier's service quality and customer relationship management.

[0126] Key indicators of partnership dimension data include the duration of cooperation and the number of cooperation projects, which are used to evaluate the supplier's cooperation stability and experience.

[0127] The key indicators of green procurement dimension data include green product procurement rate. The green product procurement rate is calculated as the proportion of green products or raw materials purchased by suppliers to total procurement.

[0128] The key indicators of green manufacturing dimension data include production process energy consumption and production process emissions. Production process energy consumption requires calculating the amount of energy consumed to produce a unit of product or provide a unit of service. Production process emissions require calculating the amount of pollutants such as wastewater, waste gas, and waste residue generated during the production process.

[0129] The key indicators of green management dimension data include green product development rate, which is calculated as the proportion of green products or raw materials developed by suppliers to the total developed products.

[0130] The key indicators of the energy conservation and emission reduction measures dimension data include energy consumption reduction rate and pollutant emission reduction rate. The energy consumption reduction rate is the ratio of energy consumption before and after improvement, and the pollutant emission reduction rate is the ratio of pollutants before and after improvement.

[0131] The key indicators of waste treatment dimension data include waste resource utilization rate. The waste resource utilization rate is calculated as the proportion of waste volume that is recycled to the total waste volume. Green and environmentally friendly helps to evaluate the impact on the environment and coordinate and optimize the economic and social benefits of the enterprise.

[0132] Step S32: Use principal component analysis or random forest algorithm to screen out typical features in each dimension and generate a typical feature subset.

[0133] Through principal component analysis (PCA) or random forest, feature extraction and dimensionality reduction are performed on data of all dimensions to select the most representative features, i.e. typical features, such as product qualification rate, on-time order delivery rate, average purchase price, after-sales service response time, cooperation duration, green product purchase rate, production process energy consumption, green product development rate, energy consumption reduction rate, waste resource utilization rate, etc., to ensure the relevance and simplicity of the features. The selected features provide a comprehensive supplier evaluation indicator. This step ensures the efficiency and effectiveness of the model and avoids overfitting problems.

[0134] Step S33: Use a convolutional neural network to perform nonlinear combination and weight assignment on the selected typical feature subsets, use historical data sets to train the model, and adjust the network structure and parameters to output a multi-dimensional feature vector for the supplier. The multi-dimensional feature vector for product management includes quality, delivery and cost dimension feature vectors, and the multi-dimensional feature vector for green environmental information includes green procurement, green manufacturing, green management, energy-saving and emission reduction measures, and waste treatment dimension feature vectors.

[0135] The convolutional neural network (CNN) is used to perform nonlinear combination and weight distribution on the selected features. The model training uses historical data sets, and by adjusting the network structure and parameters, a multi-dimensional feature vector is output, which can more carefully and accurately depict the full-dimensional feature vector of the supplier, and comprehensively evaluate the supplier based on the multi-dimensional feature vector output by the deep learning model. This step provides a comprehensive supplier evaluation index, which helps enterprises make more informed decisions when selecting suppliers.

[0136] Step S34: Construct a full-dimensional portrait of the supplier based on the multi-dimensional feature vector of product management.

[0137] These indicators reflect the supplier's quality control capabilities, delivery efficiency, cost control and service quality. Enterprises can use these comprehensive indicators to more comprehensively understand the supplier's comprehensive capabilities and help them select suitable suppliers more objectively.

[0138] Step S40: Perform green labeling processing based on the multi-dimensional feature vector of green environmental protection information to obtain a full-dimensional portrait of the supplier with a green label.

[0139] Step S40 specifically includes: Step S41: using a clustering algorithm to classify suppliers into different green levels based on the multi-dimensional feature vectors of green environmental protection information.

[0140] Data can be indexed based on basic information such as supplier name, address and contact information, and distributed crawler technology can be used to crawl green environmental protection information of each supplier in parallel. A database can be established to store the supplier's comprehensive green environmental protection information data.

[0141] If the multi-dimensional feature vector has been generated, the multi-dimensional feature vector of the supplier is green-labeled, and the green label is defined according to standards in terms of environmental protection, energy conservation, and sustainable development.

[0142] After the multi-dimensional feature vector is generated, it is compared with environmental protection standards such as ISO 14001 to define green labels, and the Kmeans clustering algorithm is used to cluster suppliers. According to the clustering results, suppliers are divided into different green levels to obtain supplier grouping information of different categories. The optimal number of clusters is determined by the elbow rule, silhouette coefficient and other methods. For example, the number of clusters is set to 3, and the optimal number of clusters is selected through SSE (sum of squared errors) evaluation, so that suppliers are divided into three green levels: high, medium and low.

[0143] The common characteristics of suppliers in each category are analyzed based on the clustering results. For example, suppliers with a high green rating may generally adopt advanced energy-saving and emission-reduction technologies and develop green products.

[0144] In other embodiments, a classification model is constructed using a decision tree model, and the classification model is trained using supplier data that has been annotated with green labels. For new suppliers, a multi-dimensional feature vector of the new supplier's green environmental information is output by the trained classification model to obtain a predicted green grade.

[0145] A classification model can be constructed through a decision tree model based on the green grades and multi-dimensional feature vectors of old suppliers (i.e. suppliers that have completed green grading through clustering algorithms). The multi-dimensional feature vectors of the new suppliers' green environmental information are input into the decision tree model to output the predicted green grade, which makes it easier to grade the green grades of subsequent new suppliers.

[0146] Use decision tree algorithms (such as the C5 algorithm) to build a classification model, select features with high information gain such as energy consumption rate, waste recycling rate, etc. as nodes, and generate decision rules to predict the green level of new suppliers. This step involves selecting effective classification attributes and pruning decision trees to optimize the performance and accuracy of the model. The multi-dimensional feature vector of each supplier's green environmental information will be input into the decision tree, and the model will output the green level label of the supplier.

[0147] The classification model is trained using the annotated green label data. During the model training process, the cross-validation method is used to optimize the decision tree structure to avoid overfitting and improve the accuracy of model prediction. For new suppliers, their feature data is input into the trained decision tree model, and the model will output the predicted green level.

[0148] Step S42: Use fuzzy reasoning technology to rate the greenness of suppliers.

[0149] Use fuzzy reasoning technology to build a green rating model and convert green grades into specific green ratings. If the supplier's green rating label has been determined by the decision tree model, use fuzzy reasoning technology to evaluate the supplier's greenness. Apply fuzzy logic technology to refine the decision tree results to obtain each supplier's green rating. Specifically, by building a fuzzy reasoning system and defining membership functions, the green rating is converted into a specific green rating, such as subdividing the "high green rating" into 85-100 points, providing a more accurate quantitative evaluation of the supplier's environmental performance.

[0150] Step S43: Integrate the green grade and the green degree score into a green label, and label the supplier with the green label.

[0151] Step S44: A full-dimensional portrait of the supplier with a green label is formed based on the multi-dimensional feature vector of the green environmental information and the green label, and the performance and score of each supplier in each dimension are recorded and displayed in detail to determine the green image of the supplier, that is, to mark such supplier as a green supplier. In other words, the supplier can be divided into green suppliers and non-green suppliers through the full-dimensional portrait of the supplier with a green label, providing a basis for corporate green procurement.

[0152] This embodiment specifically uses clustering, decision trees and fuzzy reasoning technology to green label suppliers, and by integrating the green grades and green degree scores of suppliers after green labeling, a full-dimensional portrait of suppliers with green labels is constructed, which can be used for a variety of business applications, such as supply chain management, green procurement decision support, etc. These detailed green portrait data provide a scientific basis for enterprises in supply chain management and green procurement decisions, enabling enterprises to more accurately identify and select suitable green suppliers, thereby optimizing resource allocation and improving the overall environmental protection level of the supply chain. Through this series of data analysis and model application, enterprises can not only improve environmental protection standards, but also establish a green brand image in the market and enhance competitiveness.

[0153] The present invention can apply the full-dimensional portrait of the supplier when there is no need to select a green supplier. When there is a green demand, the full-dimensional portrait of the supplier with a green label can be applied. The unified management of the two can improve the efficiency of supply chain management.

[0154] By collecting data from multiple dimensions through distributed crawler technology and cleaning, fusion and time series analysis of the data, the validity of the data can be ensured. By evaluating the credibility of the data, the reliability and quality of the data can be ensured. By using deep learning models, comprehensive evaluations of suppliers from multiple dimensions and building a full-dimensional portrait of the supplier, we can have a more comprehensive understanding of the comprehensive capabilities of the supplier and help companies choose suitable suppliers more objectively.

[0155] At the same time, the present invention also performs green labeling on the multi-dimensional feature vectors of the green environmental protection information of the supplier, constructs a full-dimensional portrait of the supplier with a green label, so as to classify the suppliers into green suppliers and non-green suppliers, and specifically performs green labeling on the suppliers through clustering, decision tree and fuzzy reasoning technology, and integrates the green grade and green degree score of the supplier after green labeling to display the performance and score of each supplier in various dimensions, providing a basis for the green procurement of the enterprise. In addition, the present invention uniformly manages the full-dimensional portrait of the supplier and the full-dimensional portrait of the supplier with a green label, which can promote the refinement and greening of the supply chain management, further help the enterprise optimize the supplier resources, enable the present invention to more comprehensively and objectively display the types and performance of suppliers, improve the efficiency of supplier management, and promote the sustainable development strategy of the enterprise.

[0156] Embodiment 2

[0157] The embodiment of the present invention also provides a supplier intelligent portrait construction system based on deep learning, the system comprising:

[0158] The processing module is used to obtain the initial data set from multiple data sources, and clean, merge and perform time series analysis on the initial data set to obtain the supplier's full life cycle quality data set, where the initial data set includes data on product management and green environmental protection information;

[0159] The credibility assessment module is used to conduct credibility assessment on the supplier's full life cycle quality data set and label each data point with a credibility label to obtain a supplier quality data set with a credibility label;

[0160] The product management portrait construction module is used to extract supplier features from multiple dimensions based on the supplier quality data set with credibility labels, and screen out typical features in each dimension through a feature selection algorithm and generate a typical feature subset. The typical feature subset is nonlinearly combined and weighted by a deep learning model to output a multi-dimensional feature vector of the supplier. The multi-dimensional feature vector includes a multi-dimensional feature vector in product management and a multi-dimensional feature vector in green environmental protection information. A full-dimensional portrait of the supplier is constructed based on the multi-dimensional feature vector in product management.

[0161] The green environmental protection portrait construction module is used to perform green labeling processing based on the multi-dimensional feature vectors of green environmental protection information to obtain a full-dimensional portrait of the supplier with a green label.

[0162] The beneficial effects of the deep learning-based supplier intelligent portrait construction system of the embodiment of the present invention are equivalent to the beneficial effects of the deep learning-based supplier intelligent portrait construction method mentioned above, and will not be elaborated here.

[0163] Furthermore, the processing module includes:

[0164] A data cleaning submodule is used to collect data on the product management and green environmental protection information from multiple data sources, and pre-process all collected data through a data cleaning algorithm to remove duplicate, erroneous and invalid data to obtain a supplier quality data set after preliminary cleaning;

[0165] The data fusion submodule is used to extract the key features of the supplier quality data set by using a feature engineering method, and to fuse the same data from different sources using a multi-source data fusion algorithm according to the key features, unify the data format and unit through data standardization, and fill in the missing values ​​using data completion technology to obtain a fused supplier full-dimensional data set with consistency and integrity;

[0166] The time series analysis submodule is used to identify the time characteristics and change trends of the supplier's full-dimensional data set using a time series analysis method, and use an interpolation algorithm to repair data missing and inconsistency problems to obtain a supplier's full life cycle quality data set with time series integrity.

[0167] Furthermore, the credibility assessment module includes:

[0168] The consistency scoring submodule is used to score the consistency of the supplier's full life cycle quality data set based on the data consistency scoring model. If the consistency score is lower than the threshold, the data source is rechecked;

[0169] The abnormal data identification submodule is used to identify potential abnormal data in the supplier's full life cycle quality data set through anomaly detection algorithms and filter out abnormal data;

[0170] The accuracy scoring submodule is used to use the data accuracy scoring algorithm to perform accuracy scoring on the screened supplier's full life cycle quality data set to obtain the accuracy index of each piece of data;

[0171] The integrity scoring submodule is used to use the data integrity scoring model to perform integrity scoring on the selected supplier's full life cycle quality data set, determine whether the information in the selected supplier's full life cycle quality data set is comprehensive, and verify the original data source again to supplement and improve the information for the incomplete data found;

[0172] The comprehensive scoring submodule is used to comprehensively score the consistency, accuracy and completeness, and uses the weighted average method to perform a comprehensive credibility score on each data in the supplier's full life cycle quality data set after screening;

[0173] The credibility label submodule is used to determine whether the comprehensive credibility score of each data in the supplier's full life cycle quality data set exceeds a preset threshold. If so, a credibility label is added to each data that exceeds the preset threshold to obtain a supplier quality data set with a credibility label.

[0174] Furthermore, the product management portrait building module includes:

[0175] The feature extraction submodule is used for supplier quality data sets with credibility labels, including supplier quality data on product management and green environmental protection information. Multidimensional feature extraction technology is used to extract multi-dimensional supplier features. The supplier quality data on product management includes quality, delivery and cost dimensions, and the supplier quality data on green environmental protection information includes green procurement, green manufacturing, green management, energy-saving and emission reduction measures and waste treatment dimensions.

[0176] The feature screening submodule is used to screen out typical features in each dimension and generate a typical feature subset using principal component analysis or random forest algorithm;

[0177] The multi-dimensional feature vector output submodule is used to use a convolutional neural network to perform nonlinear combination and weight allocation on the selected typical feature subsets, use historical data sets to train the model, and adjust the network structure and parameters to output the multi-dimensional feature vectors of suppliers. The multi-dimensional feature vectors of product management include quality, delivery and cost dimension feature vectors, and the multi-dimensional feature vectors of green environmental information include green procurement, green manufacturing, green management, energy-saving and emission reduction measures, and waste treatment dimension feature vectors;

[0178] The first construction submodule is used to construct a full-dimensional portrait of the supplier based on the multi-dimensional feature vector of product management.

[0179] Furthermore, the green environmental protection portrait construction module includes:

[0180] The clustering submodule is used to classify suppliers into different green levels using clustering algorithms based on the multi-dimensional feature vectors of green environmental information;

[0181] Output green degree scoring submodule, used to score the green degree of suppliers using fuzzy reasoning technology;

[0182] A green label submodule, for integrating the green grade and the green degree score into a green label, and labeling the supplier with the green label;

[0183] The second construction submodule is used to form a full-dimensional portrait of the supplier with a green label based on the multi-dimensional feature vector of the green environmental protection information and the green label, and record and display the performance and score of each supplier in each dimension in detail.

[0184] Embodiment 3

[0185] An embodiment of the present invention further provides a computer device, comprising:

[0186] one or more processors;

[0187] A storage device for storing one or more programs;

[0188] When one or more programs are executed by one or more processors, one or more processors implement the above-mentioned method for constructing a supplier intelligent portrait based on deep learning.

[0189] The beneficial effects of the computer device of an embodiment of the present invention are equivalent to the beneficial effects of the above-mentioned method for constructing a supplier intelligent portrait based on deep learning, and will not be elaborated here.

[0190] An embodiment of the present invention also provides a storage medium on which a computer program is stored, and when the program is processed, the above-mentioned method for constructing a supplier intelligent portrait based on deep learning is implemented.

[0191] The beneficial effects of the storage medium of the present invention are equivalent to the beneficial effects of the above-mentioned method for constructing a supplier intelligent portrait based on deep learning, and will not be elaborated here.

[0192] An embodiment of the present invention further provides a computer program product, including computer program instructions, wherein the computer program instructions enable a computer to execute any one of the above methods.

[0193] The beneficial effects of the computer program product of the present invention are equivalent to the beneficial effects of the above-mentioned method for constructing a supplier intelligent portrait based on deep learning, and will not be elaborated here.

[0194] Figure 6A schematic block diagram of an example computer device 600 that can be used to implement an embodiment of the present invention is shown. The computer device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The computer device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0195] like Figure 6 As shown, the computer device 600 includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the device can also be stored. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0196] A number of components in the computer device 600 are connected to the I / O interface 605, including: an input unit 606, such as a keyboard, a mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a storage unit 608, such as a disk, an optical disk, etc.; and a communication unit 609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 609 allows the computer device 600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0197] The computing unit 601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 601 performs the various methods and processes described above, such as a method for building a supplier intelligent profile based on deep learning. For example, in some embodiments, the method for building a supplier intelligent profile based on deep learning may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the computer device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the computing unit 601, one or more steps of the method for building a supplier intelligent profile based on deep learning described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured in any other appropriate manner (for example, by means of firmware) to execute a method for constructing a supplier intelligent portrait based on deep learning.

[0198] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. Implemented in one or more computer programs, the one or more computer programs can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a dedicated or general programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0199] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0200] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0201] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0202] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0203] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0204] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in the present invention can be achieved, and this document does not limit this.

[0205] It should be understood that although the steps in the flowchart of the accompanying drawings are shown in sequence according to the instructions of the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise clearly stated in this document, the execution of these steps is not strictly limited in order and can be performed in other orders.

[0206] Moreover, at least part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The order of execution is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0207] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for constructing supplier intelligent portrait based on deep learning, characterized in that: The method comprises: Acquire an initial data set from multiple data sources, and clean, merge, and perform time series analysis on the initial data set to obtain a supplier's full life cycle quality data set, wherein the initial data set includes data on product management and green environmental protection information; Conducting a credibility assessment on the supplier's full life cycle quality data set, and labeling each data point with a credibility label to obtain a supplier quality data set with a credibility label; Extract supplier features from multiple dimensions based on the supplier quality data set with credibility labels, screen out typical features in each dimension through a feature selection algorithm and generate a typical feature subset, perform nonlinear combination and weight assignment on the typical feature subset through a deep learning model, and output a multi-dimensional feature vector of the supplier, the multi-dimensional feature vector includes a multi-dimensional feature vector in product management and a multi-dimensional feature vector in green environmental protection information, and construct a full-dimensional portrait of the supplier based on the multi-dimensional feature vector in product management; Green labeling is performed based on the multi-dimensional feature vector of the green environmental protection information to obtain a full-dimensional portrait of the supplier with a green label.

2. The method for constructing a supplier intelligent portrait based on deep learning according to claim 1, characterized in that: The specific steps of obtaining the initial data set from multiple data sources, and performing cleaning, fusion and time series analysis on the initial data set to obtain the supplier's full life cycle quality data set include: Collect data on product management and green environmental protection information from multiple data sources, and pre-process all collected data through a data cleaning algorithm to remove duplicate, erroneous and invalid data to obtain a supplier quality data set after preliminary cleaning; The key features of the supplier quality data set are extracted by feature engineering methods. The same data from different sources are fused using a multi-source data fusion algorithm based on the key features. The data format and unit are unified through data standardization. Missing values ​​are filled using data completion technology to obtain a consistent and complete fused supplier full-dimensional data set. The time series analysis method is used to identify the time characteristics and change trends of the supplier's full-dimensional data set, and the interpolation algorithm is used to repair data missing and inconsistency problems, so as to obtain a supplier's full life cycle quality data set with temporal integrity.

3. The method for constructing a supplier intelligent portrait based on deep learning according to claim 1, characterized in that: The specific steps of performing credibility assessment on the supplier's full life cycle quality data set and labeling each data point with a credibility label to obtain the supplier quality data set with a credibility label include: Performing consistency scoring on the supplier's full life cycle quality data set based on a data consistency scoring model, and rechecking the data source if the consistency score is lower than a threshold; Identify potential abnormal data in the supplier's full life cycle quality data set through an anomaly detection algorithm, and filter out the abnormal data; Using a data accuracy scoring algorithm to score the accuracy of the supplier's full life cycle quality data set after screening, to obtain an accuracy index for each piece of data; Use the data integrity scoring model to score the integrity of the supplier's full life cycle quality data set after screening, determine whether the information in the supplier's full life cycle quality data set after screening is comprehensive, and verify the original data source again for the incomplete data found to supplement and improve the information; Combining the consistency score, the accuracy score and the completeness score, a weighted average method is used to perform a comprehensive credibility score on each data in the screened supplier's full life cycle quality data set; Determine whether the comprehensive credibility score of each data in the supplier's full life cycle quality data set exceeds a preset threshold. If so, add a credibility label to each data exceeding the preset threshold to obtain a supplier quality data set with a credibility label.

4. The method for constructing a supplier intelligent profile based on deep learning according to claim 1, characterized in that: The supplier quality data set with credibility labels is used to extract supplier features from multiple dimensions, and the typical features in each dimension are screened out through a feature selection algorithm to generate a typical feature subset. The typical feature subset is nonlinearly combined and weighted by a deep learning model to output a multidimensional feature vector of the supplier. The multidimensional feature vector includes a multidimensional feature vector in product management and a multidimensional feature vector in green environmental protection information. The specific steps of constructing a full-dimensional portrait of the supplier based on the multidimensional feature vector in product management include: The supplier quality data set with credibility labels includes supplier quality data on product management and green environmental protection information, and multi-dimensional supplier features are extracted using multi-dimensional feature extraction technology, wherein the supplier quality data on product management includes quality, delivery and cost dimensions, and the supplier quality data on green environmental protection information includes green procurement, green manufacturing, green management, energy-saving and emission reduction measures and waste treatment dimensions; Use principal component analysis or random forest algorithm to screen out typical features in each dimension and generate a typical feature subset; A convolutional neural network is used to perform nonlinear combination and weight allocation on the selected typical feature subsets, and a historical data set is used to train the model. By adjusting the network structure and parameters, a multi-dimensional feature vector of the supplier is output. The multi-dimensional feature vector of product management includes quality, delivery and cost dimension feature vectors, and the multi-dimensional feature vector of green environmental information includes green procurement, green manufacturing, green management, energy-saving and emission reduction measures, and waste treatment dimension feature vectors; A full-dimensional portrait of the supplier is constructed based on the multi-dimensional feature vectors of the product management.

5. The method for constructing a supplier intelligent portrait based on deep learning according to claim 4, characterized in that: The specific steps of performing green labeling processing according to the multi-dimensional feature vector of the green environmental protection information to obtain a full-dimensional portrait of the supplier with a green label include: According to the multi-dimensional feature vector of the green environmental protection information, the suppliers are divided into different green levels by using a clustering algorithm; Use fuzzy reasoning technology to rate suppliers’ greenness; Integrate the green grade and the green degree score into a green label, and label the supplier with the green label; A full-dimensional portrait of the supplier with a green label is formed based on the multi-dimensional feature vector of the green environmental protection information and the green label, and the performance and score of each supplier in each dimension are recorded and displayed in detail.

6. The method for constructing a supplier intelligent portrait based on deep learning according to claim 5, characterized in that: Use the decision tree model to build a classification model, and use the supplier data marked with green labels to train the classification model; For new suppliers, the multi-dimensional feature vector of the new supplier's green environmental information is output by the trained classification model to obtain a predicted green grade.

7. A supplier intelligent portrait construction system based on deep learning, characterized in that: The system comprises: A processing module is used to obtain an initial data set from multiple data sources, and clean, merge and perform time series analysis on the initial data set to obtain a supplier's full life cycle quality data set, wherein the initial data set includes data on product management and green environmental protection information; A credibility evaluation module is used to perform credibility evaluation on the supplier's full life cycle quality data set and label each data point with a credibility label to obtain a supplier quality data set with a credibility label; A product management portrait construction module is used to extract supplier features from multiple dimensions based on the supplier quality data set with credibility labels, and screen out typical features in each dimension through a feature selection algorithm and generate a typical feature subset, and perform nonlinear combination and weight assignment on the typical feature subset through a deep learning model to output a multi-dimensional feature vector of the supplier, wherein the multi-dimensional feature vector includes a multi-dimensional feature vector of product management and a multi-dimensional feature vector of green environmental protection information, and a full-dimensional portrait of the supplier is constructed based on the multi-dimensional feature vector of product management; The green environmental protection portrait construction module is used to perform green labeling processing based on the multi-dimensional feature vector of the green environmental protection information to obtain a full-dimensional portrait of the supplier with a green label.

8. A computer device, characterized in that: include: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising computer program instructions, wherein the computer program instructions enable a computer to execute the method according to any one of claims 1 to 6.

Images