Being the core of the off-site commercial banking regulation system, the off-site banking regulation system play a key role in the entire banking regulation system. In this paper, we regard the off-site regulation as a kind of pattern recognition, and we present a new model based on Self-Organizing Feature Map neural networks. Using this model, we classified commercial bank from the practice data and made a corresponding analysis of the result. The effectiveness of the proposal approach is clearly demonstrated by the experiment results.