An improved method of credit counting was introduced, which used the interval distribution of commodity prices to rate success trades and used different deducting coefficients to rate failed trades according to different credit rating. A risk evaluation approach for online transactions was proposed through analyzing the credit risks. It was based on the improved credit counting method. The approach could evaluate the risk of current trades according to the trading his-tory and current trading prices. The simulation experiments show that the improved credit counting method is preferable on anti credit hyping and anti periodical cheating, and the risk evaluation method is relatively accurate and effective when it is used to analyze and forecast risk in the transactions of e-commerce.