An effective heuristic for developing hybrid feature selection in high dimensional and low sample size datasets.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• High-dimensional datasets with low sample sizes (HDLSS) present challenges in feature selection.
• Accurate feature selection is critical for drug development and disease diagnostics in bioinformatics.
• Identifying optimal features in HDLSS remains a significant challenge.

Purpose of the Study:

• To develop an effective feature selection method for HDLSS data.
• To address the challenge of identifying informative features while discarding irrelevant ones.
• To improve prediction model performance using selected features.

Main Methods:

• A novel feature selection method combining gradual permutation filtering and a heuristic tribrid search strategy.
• Consideration of inter-feature interactions and utilization of feature rankings during the search.
• Introduction of a new performance metric for HDLSS evaluating feature number and quality.

Main Results:

• The proposed method reduced the average number of selected features from 37.8 to 5.5.
• Prediction model performance improved from 0.855 to 0.927 using the selected features.
• Demonstrated superior performance compared to existing methods on a benchmark dataset.

Conclusions:

• The developed method effectively selects a minimal set of important features.
• Achieved high prediction performance, indicating the utility of the selected features.
• Offers a robust solution for feature selection in HDLSS contexts.