Financial distress prediction is a hot topic in both theoretical and plactical area of finance. In order to identify those companies that are possible to fall into financial distress in less than two years, a new method for financial distress prediction is proposed based on grey case-based reasoning whose feature weight vector is optimized by the genetic algorithm. Meanwhile, empirical research is used to provide some evidence. There are four key techniques in the new method, i.e. enterprise case representation for financial distress prediction, k-nearest neighbor case retrieval based on the grey similarity, combination of target case class based on the similarity weighted voting, and feature weight vector optimization based on the genetic algorithm, they have been build up. In the empirical experiment with 270 Chinese listed companies’ one-yearand two yeardata before they become Special Treatment(ST) companies, grid-search technique is utilized to determine parameter values; Leave-One-Out Cross-Validation (LOO-CV) accuracy is employed as an assessment. Experiment results indicate that this new method significantly outperforms multi discriminant analysis, Logistic regression, BP neural networks, and support vector machine.