Small and medium-sized enterprises have promoted the development of the national economy, but they often ask for loaning from banks due to the shortage of funds and mortgage assets. Therefore, it is very important for banks to formulate reasonable credit strategies. This paper introduces TOPSIS and hierarchical cluster analysis algorithm, establishes a credit decision-making model based on the strength, stability of supply and demand, reputation and anti-risk ability of enterprises, quantitatively scores the risk of each enterprise, then obtains the lending rates, and establishes an objective function through RAROC theory to obtain the loan amount when the bank’s profit is the maximum. In addition, the missing indicators of enterprises without credit records can be predicted by BP neural network prediction model. This paper studies and solves the credit decision-making problems of small and medium-sized enterprises, and provides a reference for banks to put forward more accurate credit strategies.