Palmitoylation is an important hydrophobic protein modification activity that participates many cellular processes, including signaling, neuronal transmission, membrane trafficking and so on. So it is an important problem to identify palmitoylated proteins and the corresponding sites. Comparing with the expensive and time-consuming biochemical experiments, the computational methods have attracted much attention due to their good performances in predicting palmitoylation sites. In this paper, we develop a novel automated computational method to perform this work. For a sequence segment in a given protein, the encoding scheme based on the composition of k-spaced amino acid pairs (CKSAAP) is introduced, and then the support vector machine is used as the predictor. The proposed prediction model CKSAAP-Palm outperforms the existing method CSS-Palm2.0 on both cross-validation experiments and some independent testing data sets. These results imply that our CKSAAP-Palm is able to predict more potential palmitoylation sites and increases research productivity in palmitoylation sites discovery. The corresponding software can be freely downloaded from http://www.aporc.org/doc/wiki/CKSAAP-Palm.