Research on Credit Strategy Based on XGBoost Algorithm and Optimization Problem

Abstract

Small and medium-sized enterprises are the important foundation of national economic and social development, and play an important role in expanding employment, increasing income, improving people’s livelihood and national taxation. In this paper, the credit decision-making of small and medium-sized enterprises is deeply studied, and reasonable credit risk quantification results and targeted credit strategies are given. To solve the first problem, this paper establishes a multi-index scoring model based on XGBoost algorithm. According to the extracted features, the credit risk of 123 enterprises is quantitatively analyzed, and the enterprises are scored according to the quantitative results. Prioritize them by score and enterprise credit rating, take interest rate and loan amount as decision variables, and bank profit as objective function, establish a multivariate and single-objective credit strategy optimization model, and get the optimal credit strategy for each enterprise when the annual total credit is fixed. In view of the second question, this paper takes the data in Annex 1 as the training set based on the first question, and extracts the features of the relevant data in Annex 2. Using the enterprise multi-index classification scoring model based on XGBoost algorithm, the credit rating of 302 enterprises is predicted and their credit risks are quantitatively analyzed, and the enterprises are scored according to the quantitative results. Prioritize them by score and enterprise reputation grade, establish the same optimization model as the first one, and formulate specific credit rules to make the results meet the limit conditions of total credit, thus determining the interest rate and loan amount of each enterprise, and obtaining specific credit strategies. In view of the third question, this paper takes the epidemic situation in SARS-CoV-2 as an example, and comprehensively considers the impact of this sudden factor on different industries and different types of enterprises on the basis of the second question. Based on the impact of corporate profits, the data are updated, the indicators are extracted again, and the XGBoost enterprise multi-index classification scoring model is used to predict the credit rating of 302 enterprises again and quantitatively analyze the credit risk, and the enterprises are scored according to the quantitative results. Once again, they are prioritized according to the score and the reputation level of the enterprise. Use the optimization model in question 2 to adjust the corresponding credit strategy. The model established in this paper considers more indicators, and the quantitative and prediction results are reasonable. The algorithm and model can comprehensively consider the impact of unexpected factors and adjust the credit strategy in time. In addition to being applied to credit decision-making of small and medium-sized enterprises, it can also be extended to other types of financial fields, providing ideas for solving related problems.