Author(s)

Yiyang Chen

Corresponding Author

Yiyang Chen

ABSTRACT

After the 21st century, China's small and micro enterprises have sprung up like mushrooms. Due to their small size, they often do not have fixed assets to mortgage loans, so they can only apply for credit loans with high risks for banks. To solve this problem, this paper proposes a set of small and micro enterprise credit risk assessment model based on Levenberg Marquardt improved BP neural network and genetic algorithm. Through the analysis of cash flow of enterprises, the paper extracts 7 indexes of three categories, including profit margin, positive profit growth proportion, negative cash flow proportion, number of main suppliers, number of main customers, void rate and return rate. The "excellent" standard of each index is determined according to the specific index value of 425 enterprises. We use the number of indicators that meet the "excellent" standard to measure the credit risk of enterprises in eight grades. The virtual data of 1000 enterprises are simulated by Monte Carlo random method, and the rough grades are used as training samples of BP neural network. LM method is used to improve the training efficiency, and the final fitting result is good.

KEYWORDS

BP neural network, Levenberg Marquardt, Monte Carlo