Contact map prediction is of great interests for its application in fold recognition and protein 3D structure determination. In particular, we focusd on predicting non-local interactions in this paper. We employed Support Vector Machines (SVMs) as the machine learning tool and incorporated AAindex to extract correlated mutation analysis (CMA) and sequence profile (SP) features.In addition, we evaluated the effectiveness of different features for various fold classes.On average, our predictor achieved an prediction accuracy of $0.2238$ with an improvement over a random predictor of a factor $11.7$, which is better than reported studies. Our study showed that predicted secondary structure features play an important roles for the proteins containing beta structures. Models based on secondary structure features and CMA features produce different sets of predictions. Our study also suggests that models learned separately for different protein fold families may achieve better performance than a unified model.