Gene masking - a technique to improve accuracy for cancer classification with high dimensionality in microarray data.

Area of Science:

• Computational biology
• Machine learning
• Bioinformatics

Background:

• High-dimensional feature spaces often impair classification performance across various applications.
• Identifying and removing irrelevant features is crucial for improving model accuracy.

Purpose of the Study:

• To introduce a heuristic strategy, termed gene masking, for effective feature selection.
• To enhance classification accuracy by reducing the dimensionality of feature spaces.

Main Methods:

• Gene masking utilizes a binary encoded genetic algorithm for feature selection.
• The technique is integrated with classifiers during the training phase.
• It enables the discrimination of features with the most significant contribution to classification.

Main Results:

• Application on publicly available datasets demonstrated a substantial reduction in the number of features required for classification.
• High classification accuracies were maintained despite significant feature reduction.
• The method successfully isolated features with potential special significance.

Conclusions:

• Gene masking offers a powerful approach to feature selection by heuristically removing non-contributing dimensions.
• The technique demonstrably improves classifier performance, making it highly useful in data analysis.
• It aids researchers in identifying key features for further investigation.