Probabilistic Calibration and Genetic Algorithm-based Bank Credit Strategies for MSMEs and Enlightenment to Tobacco Enterprise Management

Abstract

To further strengthen the role of Micro, Small and Medium Enterprises (MSMEs) in maintaining the vitality of national economy, governments around the world introduced many special policies. They kept guiding the banking industry to increase the support for MSMEs and reduce their financing difficulties in banks. Basing on the analysis of the bank's credit strategy for small and medium-sized enterprises of similar size, this paper gives the management strategy for small and medium-sized enterprises in tobacco industry to obtain bank credit when they cannot expand their turnover. In this paper, we proposed a binary classification model-based probabilistic calibration algorithm to calculate the default probability of enterprises in the formation of risk measurement model, and found the optimal solution of credit strategy using an improved genetic algorithm. Firstly, we discovered the enterprise’s information and invoice data of 123 micro and medium-sized enterprises with existing credit ratings. We extracted several features from multiple perspectives, such as size, relationship in supply chain, profitability, performance ability, and level of development, and removed the correlations among the indicators using principal component analysis (PCA). Secondly, the retained principal components were used as covariates, and we determined the credit ratings of the firms and the probability of default using discrete variables such as the credit ratings of the firms and whether they defaulted. Finally, we substituted the probability of default into the credit risk model to calculate the loss expectation and profit expectation of the credit portfolio, and used the profit expectation of the credit portfolio as the objective function of the 0-1 programming equation to derive the credit strategy with the lowest risk exposure and the highest return basing on the genetic algorithm.