Abstract: The Mahalanobis distance discriminant method has a shortage that the weight factors can't be considered. To overcome this shortage, the rough sets theory is used to analyze and obtain weight factors. The weight coefficients are computed by analyzing the support and significance of forecasting method for the predicted object. The weighted distance discriminant models of slope stability evaluation and prediction are established by introducing weight coefficients. According to the characters of slope instability and failure, the rough set and distance discriminant models of slope stability evaluation and prediction are founded, where reasonable indexes are considered and a large set of case engineering samples are taken as raw data and training samples. The validity of rough set and distance discriminant models of slope stability evaluation and prediction have been verified through contrasting with Mahalanobis distance discriminant method, support vector machine and Bayes discriminant analysis. The rough set and distance discriminant models are applied to the slope engineering at some large water control project reservoir area in the middle reaches of Yellow River. The predicting results are according with real situation. The research results show that the rough set and distance discriminant models are reasonable for weight analysis with high prediction accuracy, which is a new and effective method for slope stability analysis and prediction.