Client credit analysis and evaluation method based on full life cycle data and AI algorithm

Abstract

The invention belongs to the technical field of project data analysis, and discloses a customer credit analysis and evaluation method based on full life cycle data and an AI algorithm. The method comprises the following steps: constructing a customer credit database according to first full life cycle data of a plurality of first engineering projects, and constructing a customer credit analysis and evaluation model by using an AI algorithm; collecting second full life cycle data of a second engineering project, and storing the second full life cycle data to the customer credit database; and performing customer credit analysis and evaluation on the second full life cycle data stored in the customer credit database by using the customer credit analysis and evaluation model to obtain a second customer credit analysis and evaluation result. According to the invention, the problems of weak default risk identification capability, poor accuracy, poor comprehensiveness and low efficiency in the prior art are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of project data analysis, and specifically relates to a customer credit analysis and evaluation method based on full life cycle data and AI algorithms. Background Art

[0002] In the current market environment, engineering design institutes often focus on "order signing," believing that revenue and profits can be achieved as long as contracts are signed and projects are completed, while ignoring the risk of client default. Therefore, in engineering project management, the client's credit status is crucial to the smooth implementation of projects and the company's risk management. With the development of artificial intelligence (AI) technology, AI models, as powerful data analysis tools, have demonstrated significant advantages in financial risk assessment and credit evaluation. However, existing technologies still have the following shortcomings: 1) While some engineering design institutes have established client credit analysis and evaluation processes within their institutional procedures, these processes are often limited to pre-contract evaluation. They lack assessment and tracking of potential default risks that may arise during project execution, making it difficult to respond to risks promptly and resulting in weak default risk identification capabilities. Engineering projects often have long lifecycles, sometimes spanning multiple cycles. However, when evaluating client credit risk, engineering design institutes focus solely on the client itself, ignoring the potential impact of other factors on changes in client credit risk. This results in poorly accurate client credit analysis and evaluation.

[0003] 2) The entire lifecycle of an engineering project involves multiple processes and departments, with data scattered across disparate systems, making effective integration difficult. Furthermore, due to a lack of effective credit assessment tools, engineering design institutes, even if they have a customer credit evaluation process within their institutional processes, often rely solely on simple financial indicators, ignoring non-financial factors such as project execution capabilities and contract performance. This lack of comprehensiveness results in an inability to timely monitor changes in customer credit risk and promptly address overdue customer credit issues, ultimately leading to bad debts in the "two funds" (reimbursement and loan) system, resulting in financial losses.

[0004] 3) Traditional design institutes lack a high level of informatization, relying on extensive manual record keeping for daily management. This creates a significant workload, high error rates, and low efficiency, making continuous tracking and management impossible. Sales, procurement, project management, and finance systems are limited by their respective development platforms. Data collection, transmission, and standards vary widely, resulting in unstable transmission. These systems operate in silos, making integration difficult, inter-system data exchange poor, and the integration of business and finance is low. In this context, managers at all levels of the company and projects receive fragmented data from various business systems. Without intuitive and effective analytical results to support decision-making, they often rely solely on empirical judgment, making it difficult to effectively assess customer credit and provide risk warnings. Summary of the Invention

[0005] In order to solve the problems of weak default risk identification ability, poor accuracy, poor comprehensiveness and low efficiency in the existing technology, the purpose of the present invention is to provide a customer credit analysis and evaluation method based on full life cycle data and AI algorithm.

[0006] The technical solution adopted in the present invention is: A customer credit analysis and evaluation method based on full life cycle data and AI algorithms includes the following steps: Build a customer credit database based on the first-life cycle data of several first-level engineering projects, and use AI algorithms to build a customer credit analysis and evaluation model; Collect the second full life cycle data of the second engineering project, and store the second full life cycle data in the customer credit database; Using the customer credit analysis and evaluation model, a customer credit analysis and evaluation is performed on the second full life cycle data stored in the customer credit database to obtain a second customer credit analysis and evaluation result.

[0007] Furthermore, based on the first full life cycle data of a number of first engineering projects, a customer credit database is constructed, and an AI algorithm is used to construct a customer credit analysis and evaluation model, including the following steps: In a third-party system, initial first full life cycle data of a plurality of first engineering projects is captured and structured converted to obtain structured first full life cycle data; Constructing a customer credit database according to the data structure of the structured first full life cycle data, and storing the structured first full life cycle data in the customer credit database; Setting a preset indicator type and preset data acquisition information, and extracting a plurality of first preset indicator contents from the structured first full life cycle data according to the preset indicator type and the preset data acquisition information; Integrate a plurality of first preset indicator contents of all structured first full life cycle data to obtain a plurality of model samples; Based on several model samples, use AI algorithms to build a customer credit analysis and evaluation model, and connect the customer credit analysis and evaluation model to the customer credit database.

[0008] Furthermore, third-party systems include business management sharing systems, project management systems, smart engineering systems, and standard financial systems.

[0009] Furthermore, the first full life cycle data includes first customer financial data, first contract data, first project management data, and first external credit data of the first engineering project; The second full life cycle data includes the second customer financial data, second contract data, second project management data and second external credit data of the second engineering project.

[0010] Furthermore, the preset indicator types include business data indicator types, financial data indicator types, customer credit indicator types, and macro-evaluation indicator types.

[0011] Furthermore, the preset data information for the business data indicator type includes customer, project name, project number, contract name, contract number, currency, contract amount, project gross profit, completed project volume / work volume, cash collection amount, non-cash collection amount, contract receivables but not collected, and contract overdue receivables; The preset data information for financial data indicator types includes customer, project name, project number, contract name, contract number, currency, project type, project status, invoicing date, invoicing amount, contract asset amount, and contract asset entry date. The preset data for the merchant credit indicator type includes customer name, company establishment date, registered capital, company nature, number of litigation disputes, third-party credit rating, third-party risk analysis, balance sheet data, income statement data, and cash flow statement data. The preset data information of the macro-evaluation indicator type includes GDP information, macro-evaluation information of large-scale enterprises, PPI information, economic information, steel industry information, and currency and interest rate information.

[0012] Furthermore, the preset indicators of the business data indicator type include the length of cooperation, the proportion of contract value in the past three years, the proportion of contract quantity, the total amount of contract receivables, the average profit margin of projects in the past three years, the customer cash payment ratio in the past year, customer overdue receivables, and customer overdue rate; The preset indicators of the financial data indicator type include the aging distribution of customer accounts receivable, the aging distribution of customer contract assets, the balance of customer accounts receivable by different project types, the balance of customer contract assets by different project types, the balance of customer accounts receivable by different project statuses, and the balance of customer contract assets by different project statuses. The preset indicators of the merchant credit index type include registered capital, years of operation, enterprise nature, return on net assets, operating profit margin, return on total assets, debt-to-asset ratio, current ratio, velocity ratio, accounts receivable turnover, current asset turnover, total asset turnover, sales growth rate, and sales profit growth rate; The preset indicators of the macro-evaluation indicator type include GDP at current prices, secondary industry GDP at current prices, industrial GDP at current prices, cumulative value of operating income of industrial enterprises above designated size, cumulative value of total profits of industrial enterprises above designated size, year-on-year value added of ferrous metal smelting and rolling processing industry, month-on-month PPI of all industrial products, year-on-year PPI of metallurgical industry, month-on-month PPI of ferrous metal smelting and rolling processing industry, preliminary value of industrial production of CICC CMI index, preliminary value of domestic demand of CICC CMI index, preliminary value of external demand of CICC CMI index, purchasing managers' index of steel industry, comprehensive steel price index, iron ore index, total profit of ferrous metal smelting and rolling processing industry, cumulative year-on-year value of operating income of ferrous metal smelting and rolling processing industry, asset-liability ratio of ferrous metal smelting and rolling processing industry, profitability of steel mills, one-year LPR interest rate, year-on-year balance of medium- and long-term loans to heavy industry in RMB and foreign currencies, and cumulative value of new medium- and long-term loans to industry in RMB and foreign currencies by financial institutions.

[0013] Furthermore, based on several model samples, an AI algorithm is used to construct a customer credit analysis and evaluation model, and the customer credit analysis and evaluation model is connected to the customer credit database, including the following steps: Set a corresponding preset customer credit analysis evaluation label for each model sample, and divide it into several model training samples and several model testing samples in a ratio of 7:3; Use AI algorithms to build an initial customer credit analysis and evaluation model, and input several model training samples for optimization training to obtain an optimized customer credit analysis and evaluation model; Input several model test samples, conduct model testing, and obtain the model performance value of the optimized customer credit analysis and evaluation model; If the model performance value is greater than the performance value threshold, the final customer credit analysis and evaluation model is output and connected to the customer credit database; otherwise, optimization training continues.

[0014] Furthermore, the AI ​​algorithms include the XGBClassifier-LSTM algorithm, the SVC-RBF algorithm, and the CatBoostClassifier algorithm.

[0015] Furthermore, using the customer credit analysis and evaluation model, performing customer credit analysis and evaluation on the second full life cycle data stored in the customer credit database to obtain a second customer credit analysis and evaluation result includes the following steps: Extracting the second preset indicator content of the second full life cycle data based on the preset indicator type and the corresponding preset data acquisition information; Based on the second preset indicator content of the second full life cycle data, a customer credit analysis and evaluation model is used to make a prediction to obtain a prediction result of the second customer's overdue default rate; Based on the preset scoring mechanism, according to the overdue default rate prediction result of the second customer, the customer credit analysis and evaluation is performed to obtain the second customer credit analysis and evaluation result.

[0016] The beneficial effects of the present invention are: The present invention provides a customer credit analysis and evaluation method based on full lifecycle data and AI algorithms. This method focuses on changes in customer credit throughout the lifecycle of a project, promptly identifies customer default risks, improves the ability to identify default risks, monitors in real time factors such as the customer's own and overall industry transaction conditions, and comprehensively considers the impact of multiple factors on changes in customer credit risk, thereby improving the accuracy of customer credit analysis and evaluation. By integrating full lifecycle data from pre-sales, execution, to project completion, including multi-dimensional information such as financial data, contract data, project management data, external credit data, and macro-evaluation indicators, this method avoids the limitations of traditional credit assessment methods that rely on single-dimensional data, improves the comprehensiveness and accuracy of the assessment, and can fully reflect the customer's credit status. The customer credit analysis and evaluation model constructed based on the AI ​​algorithm dynamically updates the customer credit analysis and evaluation results based on data, ensuring the timeliness and accuracy of the customer credit analysis and evaluation results. Automatic daily data capture is achieved based on the customer credit database. The customer credit analysis and evaluation model can automatically process and analyze large amounts of data, extract features, reduce manual intervention, and improve credit assessment efficiency. This method realizes automatic analysis and evaluation of customer credit risks throughout the entire project lifecycle and real-time monitoring, significantly reducing labor and time costs and lowering operating costs.

[0017] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flowchart of the customer credit analysis and evaluation method based on full life cycle data and AI algorithms in the present invention. DETAILED DESCRIPTION

[0019] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments.

[0020] Example: like Figure 1 As shown, this embodiment provides a customer credit analysis and evaluation method based on full life cycle data and AI algorithms, including the following steps: S1: Based on the first full life cycle data of several first engineering projects, a customer credit database is constructed, and an AI algorithm is used to construct a customer credit analysis and evaluation model, including the following steps: S1-1: In a third-party system, initial first full life cycle data of a plurality of first engineering projects is captured and structured converted to obtain structured first full life cycle data; Third-party systems include business management sharing systems, project management systems, smart engineering systems, and standard financial systems; The first full life cycle data includes the first customer financial data, first contract data, first project management data, and first external credit data of the first engineering project; The steps include: S1-1-1: In a third-party system, capture the initial first life cycle data of several first engineering projects, perform field cleaning and code standardization, and obtain several standardized first life cycle data; Missing value processing: Numerical fields are filled with mean / median or outlier interpolation; categorical fields are processed with "missing" labels; Category field encoding: One-hot encoding is used for category fields such as "project type" and "enterprise nature". Ordinal encoding is used for fields with hierarchical relationships (such as rating level). Amount standardization: All amount fields (such as contract amount and payment amount) are converted to a unified currency (such as RMB) by currency and normalized using Z-score or min-max. S1-1-2: Generate time-derived features on a number of standardized first full life cycle data to obtain a number of feature-generated first full life cycle data; Years of operation = current date - date of establishment of the enterprise; Aging distribution: divide accounts receivable into aging ranges and generate percentages; Trend characteristics: Extract trend slope and periodic volatility from time series data such as workload and collection amount; Timeliness indicators: such as the number of days since the last overdue payment; S1-1-3: Establish a macro-evaluation data mapping mechanism, and map the first full life cycle data generated by a number of features according to the macro-evaluation data mapping mechanism to obtain a number of mapped first full life cycle data; Establish an industry-indicator mapping table, mapping macroeconomic evaluation data such as the steel price index and the Producer Price Index (PPI) to each sample according to the "client industry classification"; set a sliding window to smooth the macroeconomic evaluation data on a monthly or quarterly basis (such as a rolling mean) to avoid sudden changes that interfere with the model; S1-1-4: Perform data enhancement on the mapped first lifecycle data to obtain structured first lifecycle data; Data augmentation includes: Synthetic Minority Over-sampling Technique (SMOTE), which uses interpolation in feature space to synthesize "virtual high-risk customers" to expand minority class samples, preserve the original data distribution structure, and avoid overfitting caused by replication; Dynamic sample weight adjustment: During training, high-risk samples are given higher loss penalty weights, and Focal Loss is used instead of traditional cross entropy to focus on difficult-to-classify samples, improving the robustness of subsequent models. Abnormal sample generation: Generate perturbation samples (such as "extending payment terms" or "lowering collection rates") for historical customer data that has defaulted, simulating boundary samples to improve the boundary fitting capabilities of subsequent models; S1-2: Constructing a customer credit database based on the data structure of the structured first full life cycle data, and storing the structured first full life cycle data in the customer credit database; S1-3: Setting a preset indicator type and preset data acquisition information, and extracting a plurality of first preset indicator contents from the structured first full life cycle data according to the preset indicator type and the preset data acquisition information; The preset indicator types include business data indicator types, financial data indicator types, merchant credit indicator types, and macro-evaluation indicator types; The preset data information for business data indicator types includes customer, project name, project number, contract name, contract number, currency, contract amount, project gross profit, completed work / work volume, cash collection amount, non-cash collection amount, contract receivables but not collected, and contract overdue receivables. These reflect the customer's behavior in project execution, contract performance, and payment and collection. Amount fields are standardized using the Z-score; time series fields such as "Completed Work Volume" generate trend slopes and volatility on a monthly basis; and overdue receivables fields are used for outlier annotation and risk scoring weighting. The preset data information for financial data indicators includes customer, project name, project number, contract name, contract number, currency, project type, project status, invoicing date, invoicing amount, contract asset amount, and contract asset entry date, which is used to capture project execution compliance. The preset data for the merchant credit indicator type includes customer name, company establishment date, registered capital, company nature, number of litigation disputes, third-party credit rating, third-party risk analysis, balance sheet data, income statement data, and cash flow statement data. Preset data for macro-evaluation indicators includes gross domestic product (GDP), macro-evaluation information for enterprises above a designated size, producer price index (PPI), economic conditions, steel industry information, and currency and interest rate information. Through an industry / region / time matching mapping mechanism, macro-evaluation indicators can be dynamically embedded into client project cycles, enhancing risk identification capabilities. The preset indicators of the business data indicator type include the length of cooperation, the proportion of contract value in the past three years, the proportion of contract quantity, the total amount of contract receivables, the average profit margin of projects in the past three years, the customer cash payment ratio in the past year, customer overdue receivables, and customer overdue rate; The preset indicators of the financial data indicator type include the aging distribution of customer accounts receivable, the aging distribution of customer contract assets, the balance of customer accounts receivable by different project types, the balance of customer contract assets by different project types, the balance of customer accounts receivable by different project statuses, and the balance of customer contract assets by different project statuses. The preset indicators of the merchant credit index type include registered capital, years of operation, enterprise nature, return on net assets, operating profit margin, return on total assets, debt-to-asset ratio, current ratio, velocity ratio, accounts receivable turnover, current asset turnover, total asset turnover, sales growth rate, and sales profit growth rate; The preset indicators of the macro-evaluation indicator type include GDP at current prices, secondary industry GDP at current prices, industrial GDP at current prices, cumulative operating income of industrial enterprises above designated size, cumulative total profits of industrial enterprises above designated size, year-on-year value added of ferrous metal smelting and rolling processing industry, month-on-month PPI of all industrial products, year-on-year PPI of metallurgical industry, month-on-month PPI of ferrous metal smelting and rolling processing industry, preliminary value of industrial production of CICC Macroeconomic Momentum Index (CMI), preliminary value of domestic demand of CICC CMI index, preliminary value of external demand of CICC CMI index, purchasing managers' index of steel industry, comprehensive steel price index, iron ore index, total profits of ferrous metal smelting and rolling processing industry, cumulative year-on-year operating income of ferrous metal smelting and rolling processing industry, debt-to-asset ratio of ferrous metal smelting and rolling processing industry, profitability of steel mills, one-year loan prime rate (LPR), year-on-year balance of medium- and long-term loans to heavy industry in RMB and foreign currencies, and cumulative value of new medium- and long-term loans to industry in RMB and foreign currencies from financial institutions. S1-4: Integrate a plurality of first preset indicator contents of all structured first life cycle data to obtain a plurality of model samples; S1-5: Based on several model samples, use AI algorithms to build a customer credit analysis and evaluation model, and connect the customer credit analysis and evaluation model to the customer credit database, including the following steps: S1-5-1: Set a corresponding preset customer credit analysis evaluation label for each model sample, and divide it into several model training samples and several model testing samples in a ratio of 7:3; S1-5-2: Use AI algorithms to build an initial customer credit analysis and evaluation model, and input several model training samples for optimization training to obtain an optimized customer credit analysis and evaluation model; AI algorithms include the Extreme Gradient Boosting Classifier (XGBClassifier)-Long Short-Term Memory (LSTM) algorithm, the C-Support Vector Classification with Radial Basis Function Kernel (SVC-RBF) algorithm, and the CategoricalBoostingClassifier (CatBoostClassifier) ​​algorithm. Among them, the XGBClassifier-LSTM algorithm has strong ability to handle nonlinear relationships and fast calculation speed. The SVC-RBF algorithm is suitable for high-dimensional data and has clear decision boundaries. The CatBoostClassifier algorithm is suitable for categorical features and automatically handles missing values. S1-5-3: Input several model test samples, perform model testing, and obtain the model performance value of the optimized customer credit analysis and evaluation model; In this embodiment, the model performance value is the area under the curve (AUC), which is used to measure the performance of a system or model. The closer the AUC value is to 1, the stronger the model's discriminatory ability is. The curve is the receiver operating characteristic (ROC) curve. The XGBClassifier algorithm has the highest AUC (0.982) and is suitable for customer credit analysis and evaluation models, and can be combined with the LSTM algorithm for customer credit analysis and evaluation. S1-5-4: If the model performance value is greater than the performance value threshold, the final customer credit analysis and evaluation model is output and connected to the customer credit database; otherwise, optimization training continues; S2: Collect the second full life cycle data of the second project and store the second full life cycle data in the customer credit database; The second full life cycle data includes the second customer financial data, second contract data, second project management data and second external credit data of the second engineering project; S3: Using the customer credit analysis and evaluation model, perform customer credit analysis and evaluation on the second full life cycle data stored in the customer credit database to obtain a second customer credit analysis and evaluation result, including the following steps: S3-1: extracting the second preset indicator content of the second life cycle data based on the preset indicator type and the corresponding preset data acquisition information; In this embodiment, the second preset indicators include the business data indicator type, the number of years of cooperation, the proportion of contract value in the past three years, the proportion of contract quantity, the total amount of contract receivables, the average profit margin of projects in the past three years, and the customer cash payment ratio in the past year; Including accounts receivable with an age of less than 3 months, accounts receivable with an age of 3-6 months, accounts receivable with an age of 6-12 months, accounts receivable with an age of 1-1.5 years, accounts receivable with an age of 1.5-2 years, accounts receivable with an age of 2-2.5 years, accounts receivable with an age of 2.5-3 years, accounts receivable with an age of more than 3 years, accounts receivable balance of customer EPC projects, accounts receivable balance of customer design projects, accounts receivable balance of other types of customer projects, accounts receivable balance of customer ongoing projects, accounts receivable balance of customer projects entering the warranty period, accounts receivable balance of customer projects entering the contract closing period Balance of accounts receivable; contract assets with an aging of less than 3 months, contract assets with an aging of 3-6 months, contract assets with an aging of 6-12 months, contract assets with an aging of 1-1.5 years, contract assets with an aging of 1.5-2 years, contract assets with an aging of 2-2.5 years, contract assets with an aging of 2.5-3 years, contract assets with an aging of more than 3 years, balance of contract assets for customer EPC projects, balance of contract assets for customer design projects, balance of contract assets for other types of customer projects, balance of contract assets for customer projects in progress, balance of contract assets for customer projects entering the warranty period, and balance of contract assets for customer projects entering the final stages of contract completion; Including the merchant credit indicators such as registered capital, years of operation, nature of enterprise, return on net assets, operating profit margin, return on total assets, debt-to-asset ratio, current ratio, velocity ratio, accounts receivable turnover, current asset turnover, total asset turnover, sales growth rate and sales profit growth rate; Including the macro evaluation indicators such as GDP at current price, secondary industry GDP at current price, industrial GDP at current price, cumulative value of operating income of industrial enterprises above designated size, cumulative value of total profits of industrial enterprises above designated size, year-on-year value added of ferrous metal smelting and rolling processing industry, month-on-month PPI of all industrial products, year-on-year PPI of metallurgical industry, month-on-month PPI of ferrous metal smelting and rolling processing industry, preliminary value of industrial production of CICC CMI index, preliminary value of domestic demand of CICC CMI index, preliminary value of external demand of CICC CMI index, purchasing managers' index of steel industry, comprehensive steel price index, iron ore index, total profits of ferrous metal smelting and rolling processing industry, cumulative year-on-year value of operating income of ferrous metal smelting and rolling processing industry, asset-liability ratio of ferrous metal smelting and rolling processing industry, profitability of steel mills, 1-year LPR rate, year-on-year balance of medium- and long-term loans to heavy industry in RMB and foreign currencies, and cumulative value of new medium- and long-term loans to industry in RMB and foreign currencies from financial institutions; S3-2: Based on the second preset indicator content of the second full life cycle data, use the XGBClassifier module of the customer credit analysis and evaluation model to perform a prediction to obtain a prediction result of the second customer's overdue default rate; The second customer overdue default rate prediction result is the customer's overdue default rate in the next month, with a value range of 0-1; In addition to the XGBClassifier module, you can also use logistic regression, random forest, Light Gradient Boosting Machine (LightGBM), extreme gradient boosting machine (eXtreme GradientBoosting, XGBoost), support vector machine, and shallow neural network; S3-3: Based on the preset scoring mechanism and the predicted overdue default rate of the second customer, a credit analysis and evaluation of the customer is performed to obtain a credit analysis and evaluation result of the second customer, including the following steps: S3-3-1: Based on a preset scoring mechanism, collect the second enterprise's financial status data corresponding to the second full life cycle data, including overdue balances, contract collections, etc., and the second industry environment data, including PPI data and Purchasing Managers' Index (PMI) data; S3-3-2: Combine the second customer's overdue default rate prediction result with the second enterprise's financial status data and the second industry environment data to obtain second combined data; S3-3-3: Based on the second combined data, use the LSTM module of the customer credit analysis and evaluation model to perform customer credit analysis and evaluation to obtain a second customer credit analysis and evaluation result; The second customer credit analysis evaluation result may include a customer credit analysis score. The preset scoring mechanism is as follows: Credit score range: 300–950; Risk classification: high risk (<580), medium risk (580–700), low risk (≥700); Use explanatory tools such as Shapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) to output field-level score contributions; Each scoring result can be traced back to its input feature source, field value and its changing trend through the model structure; a "scoring explanation report" is provided, including field ranking, contribution value and recommended items; Provides interpretable explanations for each rating by calculating the "marginal contribution" of each feature to the predicted value; supports both local explanations (single customer) and global explanations (overall feature importance); Supports visual reports and field-level scoring backtracking for compliance and auditing; Risk control label output: Supports a rule enhancement mechanism that combines model output with manual risk control rules. For example, if a customer has a credit rating below B and is overdue for more than 90 days, the system will output an "extremely high risk warning." This provides credit review personnel with a decision-making support view, improving decision-making transparency and efficiency.

[0021] The present invention provides a customer credit analysis and evaluation method based on full life cycle data and AI algorithm, which focuses on the changes in customer credit throughout the life cycle of the project, timely discovers the customer's default risk, improves the ability to identify default risk, monitors the influencing factors such as the customer itself and the overall industry transaction situation in real time, and comprehensively considers the impact of multiple influencing factors on the changes in customer credit risk, thereby improving the accuracy of customer credit analysis and evaluation; by integrating the full life cycle data of the project from pre-sale, execution to project completion, including financial data, contract data, project management data, external credit data, macro-evaluation indicators and other multi-dimensional information, it avoids the limitations of traditional credit assessment methods that rely on single-dimensional data, improves the comprehensiveness and accuracy of the assessment, and can fully reflect the customer's credit status; constructing a customer credit analysis and evaluation model based on the AI ​​algorithm realizes the dynamic update of customer credit analysis and evaluation results according to data, ensuring customer credit The timeliness and accuracy of the analysis and evaluation results are enhanced, and daily data is automatically captured based on the customer credit database. The customer credit analysis and evaluation model can automatically process and analyze large amounts of data, extract features, reduce manual intervention, and improve credit assessment efficiency. It realizes automatic analysis and evaluation of customer credit risks throughout the entire project cycle and real-time monitoring, significantly reducing labor costs and time costs, and lowering operating costs. By accurately identifying high-risk customers, intervention can be made before sales, and strict credit conditions or scope of cooperation can be set for high-risk customers to avoid excessive risk-taking. During project execution or the final stage of the project, high-risk customers can be identified through real-time monitoring, and corresponding measures can be formulated to reduce the risk of default and bad debts. The customer credit analysis and evaluation model can combine macroeconomic changes to infer changes in customer credit risk, overcoming the shortcomings of traditional methods that cannot accurately predict the impact of macroeconomic changes on changes in customer credit risk of long-cycle projects.

[0022] The present invention is not limited to the above optional embodiments. Anyone can derive various other forms of products based on the teachings of the present invention. The above specific embodiments should not be construed as limiting the scope of protection of the present invention. The scope of protection of the present invention shall be based on the scope defined in the claims, and the description can be used to interpret the claims.

Claims

1. A customer credit analysis and evaluation method based on full lifecycle data and AI algorithms, characterized by: The steps include: Build a customer credit database based on the first-life cycle data of several first-level engineering projects, and use AI algorithms to build a customer credit analysis and evaluation model; Collect the second full life cycle data of the second engineering project, and store the second full life cycle data in the customer credit database; Using the customer credit analysis and evaluation model, a customer credit analysis and evaluation is performed on the second full life cycle data stored in the customer credit database to obtain a second customer credit analysis and evaluation result.

2. The customer credit analysis and evaluation method based on full life cycle data and AI algorithms according to claim 1, characterized in that: Based on the first-life cycle data of several first-level engineering projects, a customer credit database is constructed, and an AI algorithm is used to build a customer credit analysis and evaluation model, including the following steps: In a third-party system, initial first full life cycle data of a plurality of first engineering projects is captured and structured converted to obtain structured first full life cycle data; Constructing a customer credit database according to the data structure of the structured first full life cycle data, and storing the structured first full life cycle data in the customer credit database; Setting a preset indicator type and preset data acquisition information, and extracting a plurality of first preset indicator contents from the structured first full life cycle data according to the preset indicator type and the preset data acquisition information; Integrate a plurality of first preset indicator contents of all structured first full life cycle data to obtain a plurality of model samples; Based on several model samples, use AI algorithms to build a customer credit analysis and evaluation model, and connect the customer credit analysis and evaluation model to the customer credit database.

3. The customer credit analysis and evaluation method based on full life cycle data and AI algorithms according to claim 2 is characterized by: The third-party systems include business management sharing system, project management system, smart engineering system and standard financial system.

4. The customer credit analysis and evaluation method based on full life cycle data and AI algorithms according to claim 3 is characterized by: The first full life cycle data includes the first customer financial data, first contract data, first project management data and first external credit data of the first engineering project; The second full life cycle data includes second customer financial data, second contract data, second project management data and second external credit data of the second engineering project.

5. The customer credit analysis and evaluation method based on full life cycle data and AI algorithms according to claim 4 is characterized by: The preset indicator types include business data indicator types, financial data indicator types, customer credit indicator types and macro-evaluation indicator types.

6. The customer credit analysis and evaluation method based on full life cycle data and AI algorithms according to claim 5 is characterized by: The preset data information of the business data indicator type includes customer, project name, project number, contract name, contract number, currency, contract amount, project gross profit, completed project volume / work volume, cash collection amount, non-cash collection amount, contract receivables but not collected, and contract overdue receivables; The preset data information of the financial data indicator type includes customer, project name, project number, contract name, contract number, currency, project type, project status, invoicing date, invoicing amount, contract asset amount and contract asset accounting date; The preset data collection information of the merchant credit indicator type includes the customer name, enterprise establishment date, registered capital, enterprise nature, number of litigation disputes, third-party credit rating, third-party risk analysis, balance sheet item data, income statement item data, and cash flow statement item data; The preset data information of the macro-evaluation indicator type includes GDP information, macro-evaluation information of large-scale enterprises, PPI information, economic information, steel industry information, and currency and interest rate information.

7. The customer credit analysis and evaluation method based on full life cycle data and AI algorithms according to claim 6, characterized in that: The preset indicators of the business data indicator type include the length of cooperation, the proportion of contract value in the past three years, the proportion of contract quantity, the total amount of contract receivables, the average profit margin of projects in the past three years, the customer cash payment ratio in the past year, customer overdue receivables and customer overdue rate; The preset indicators of the financial data indicator type include the aging distribution of customer accounts receivable, the aging distribution of customer contract assets, the balance of customer accounts receivable by different project types, the balance of customer contract assets by different project types, the balance of customer accounts receivable by different project statuses, and the balance of customer contract assets by different project statuses; The preset indicators of the merchant credit index type include registered capital, years of operation, nature of enterprise, return on net assets, operating profit margin, return on total assets, debt-to-asset ratio, current ratio, velocity ratio, accounts receivable turnover, current asset turnover, total asset turnover, sales growth rate, and sales profit growth rate; The preset indicators of the macro-evaluation indicator types include GDP at current prices, secondary industry GDP at current prices, industrial GDP at current prices, cumulative value of operating income of industrial enterprises above designated size, cumulative value of total profits of industrial enterprises above designated size, year-on-year value added of ferrous metal smelting and rolling processing industry, month-on-month PPI of all industrial products, year-on-year PPI of metallurgical industry, month-on-month PPI of ferrous metal smelting and rolling processing industry, preliminary value of industrial production of CICC CMI index, preliminary value of domestic demand of CICC CMI index, preliminary value of external demand of CICC CMI index, purchasing managers' index of steel industry, comprehensive steel price index, iron ore index, total profit of ferrous metal smelting and rolling processing industry, cumulative year-on-year value of operating income of ferrous metal smelting and rolling processing industry, asset-liability ratio of ferrous metal smelting and rolling processing industry, profitability of steel mills, one-year LPR interest rate, year-on-year balance of medium- and long-term loans to heavy industry in RMB and foreign currencies, and cumulative value of new medium- and long-term loans to industry in RMB and foreign currencies by financial institutions.

8. The customer credit analysis and evaluation method based on full life cycle data and AI algorithms according to claim 7 is characterized by: Based on several model samples, use AI algorithms to build a customer credit analysis and evaluation model, and connect the customer credit analysis and evaluation model to the customer credit database, including the following steps: Set a corresponding preset customer credit analysis evaluation label for each model sample, and divide it into several model training samples and several model testing samples in a ratio of 7:3; Use AI algorithms to build an initial customer credit analysis and evaluation model, and input several model training samples for optimization training to obtain an optimized customer credit analysis and evaluation model; Input several model test samples, conduct model testing, and obtain the model performance value of the optimized customer credit analysis and evaluation model; If the model performance value is greater than the performance value threshold, the final customer credit analysis and evaluation model is output and connected to the customer credit database; otherwise, optimization training continues.

9. The customer credit analysis and evaluation method based on full life cycle data and AI algorithms according to claim 8, characterized in that: The AI ​​algorithms include XGBClassifier-LSTM algorithm, SVC-RBF algorithm and CatBoostClassifier algorithm.

10. The customer credit analysis and evaluation method based on full life cycle data and AI algorithm according to claim 9 is characterized by: Using the customer credit analysis and evaluation model, performing customer credit analysis and evaluation on the second full life cycle data stored in the customer credit database to obtain a second customer credit analysis and evaluation result includes the following steps: Extracting the second preset indicator content of the second full life cycle data based on the preset indicator type and the corresponding preset data acquisition information; Based on the second preset indicator content of the second full life cycle data, a customer credit analysis and evaluation model is used to make a prediction to obtain a prediction result of the second customer's overdue default rate; Based on the preset scoring mechanism, according to the overdue default rate prediction result of the second customer, the customer credit analysis and evaluation is performed to obtain the second customer credit analysis and evaluation result.

Images