Predicting hot spots in protein-protein interface by using computational methods is of vital significance as it provides crucial information for protein function research and drug design.Hot spots are not isolated, but tend to interact with other amino acid residues.Diverse neighboring residues show different influence on hot spots.Complex network method was applied to uncover such influence and then several novel features were designed to describe the diversity of environment of hot spots.Thereafter, feature analysis was carried on 75 hybrid features.The novel features are top-ranked in feature importance permutation.An optimal 58-feature set was then applied to construct a Support Vector Machine (SVM) prediction model for hot spots.