Using Rotation Forest for Protein Fold Prediction Problem: An Empirical Study

摘要

Recent advancement in the pattern recognition field has driven many classification algorithms being implemented to tackle protein fold prediction problem. In this paper, a newly introduced method called Rotation Forest for building ensemble of classifiers based on bootstrap sampling and feature extraction is implemented and applied to challenge this problem. The Rotation Forest is a straight forward extension of bagging algorithms which aims to promote diversity within the ensemble through feature extraction by using Principle Component Analysis (PCA). We compare the performance of the employed method with other Meta classifiers that are based on boosting and bagging algorithms, such as: AdaBoost.Ml, LogitBoost, Bagging and Random Forest. Experimental results show that the Rotation Forest enhanced the protein folding prediction accuracy better than the other applied Meta classifiers, as well as the previous works found in the literature.