A prediction framework based on extreme learning machine for secondary structure of protein

A prediction framework was proposed for training the secondary structure model of protein, based on a new effective learning algorithm, i.e., the extreme learning machine (ELM). Then, to merge the predicted results together better, a probability-based combining (PBC) algorithm was proposed with a Helix-post-processing (HPP) algorithm set out according to the biological features of protein's secondary structure, which will provide efficient post-processing effect on the predicted results after merging so as to improve their accuracy further. The experiments were carried out on the datasets CB513 and RS126 separately, and the predicted results showed that the accuracy of the proposed algorithms is satisfactory especially the training time that is shortened greatly.