A novel random forests-based feature selection method for microarray expression data analysis.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• High-dimensional data and redundant features pose challenges in bioinformatics.
• Effective feature selection is crucial for accurate biological data analysis.

Purpose of the Study:

• To propose a novel random forests-based feature selection method.
• To enhance classification accuracy and reduce computational time in bioinformatics research.

Main Methods:

• Stratifying feature space using random forests.
• Combining generalized sequence backward and forward searching strategies.
• Utilizing random forest variable importance scores for feature ranking.
• Employing various classifiers for feature subset evaluation.

Main Results:

• The method was tested on five microarray datasets (leukaemia, prostate, breast, nervous, DLBCL).
• Support Vector Machine (SVM) classifier achieved average accuracies of 100%, 95.24%, 85%, 91.67%, and 91.67% respectively.
• Demonstrated significant improvements in classification accuracy.
• Showcased a substantial reduction in feature selection computation time.

Conclusions:

• The proposed random forests-based method is effective for feature selection in bioinformatics.
• It offers a computationally efficient approach to improve classification accuracy.
• This method addresses the challenges of high-dimensional and redundant data in biological research.