Prediction of Protein Secondary Structure Using the Hierarchical Competitive Covering Networks Based on Quotient Space

In this paper, the algorithm of a kind of hierarchical competitive covering networks for the classification problems is proposed based on the quotient space theory, which defines granularity according to Huffman coding. Not only three classes of secondary structure but also eight classes are discussed for the protein. Instances show that this kind of networks improves the sorting ability of covering net works, the algorithm is feasible and high performing, Moreover different encoding method lead different prediction accuracy, the code contains more information of protein structure can obtain higher prediction precision.