Predicting Yeast Protein Localization Sites by a New Clustering Algorithm Based on Weighted Feature Ensemble

摘要

The analysis of protein localization sites is an important task in bioinformatics. Predicting the yeast protein localization sites is a promising domain among numerous research methods based on the yeast protein measurement data which have multiple indexes/features. In order to reflect the different contributions of those features to predicting tasks, a clustering algorithm based on weighted feature ensemble (WFE) is proposed to predict yeast protein localization sites on the basis of the gathered yeast protein localization data. WFE process firstly assigns different weights to features, and then the results are computed and presented to obtain the best outcome. Experimental results on our algorithm based on WFE and other several clustering algorithms based on the ideas of weighted features have shown that our new algorithm outperformed the other feature weighting type algorithms in accuracy and stability.