Optimizing effective fusion representation by particle swarm optimization algorithm for protein sub-nuclear location

Feature representation contains the more plentiful information of original protein sequence, the more beneficial for protein sub-nuclear localization. Inspired by this idea, this paper proposed a novel two-feature integration method, whose fusion parameter was optimized via the particle swarm optimization algorithm (PSO), for obtaining a more effective representation. Therefore, a new fusion representation, called AACPSSM, would be formed by integrating two kinds of single feature expression, amino acid composition (AAC) and position specific scoring matrix (PSSM). Due to the high dimensional characteristics of protein data, kernel linear discriminant analysis (KLDA) was used to conduct the data dimension reduction. Last, to evaluate validity of our proposed approach, a benchmark dataset and KNN classifier were used to carry out the numerical experiments. And the final Jackknife test experimental results prove that our proposed fusion representation AACPSSM largely outperforms the single one, AAC and PSSM.