Feature Selection for Optimized High-Dimensional Biomedical Data Using an Improved Shuffled Frog Leaping Algorithm.

Area of Science:

• Biomedical data analysis
• Machine learning in healthcare
• Computational biology

Background:

• High-dimensional biomedical datasets pose challenges for disease diagnosis due to numerous irrelevant features.
• Effective feature selection is crucial for improving classification accuracy and reducing computational overhead in data mining.

Purpose of the Study:

• To develop an improved feature selection algorithm for high-dimensional biomedical data.
• To maximize predictive accuracy while minimizing irrelevant features.
• To reduce computational complexity in data mining for molecular diagnosis.

Main Methods:

• An improved shuffled frog leaping algorithm incorporating a chaos memory weight factor, absolute balance group strategy, and adaptive transfer factor.
• Feature subset exploration to identify optimal feature sets.
• Evaluation using the K-nearest neighbor classifier with comparative analysis against genetic algorithms, particle swarm optimization, and the standard shuffled frog leaping algorithm.

Main Results:

• The proposed improved shuffled frog leaping algorithm demonstrated enhanced identification of relevant feature subsets.
• Significant improvements in classification accuracy were observed compared to existing methods.
• The algorithm effectively reduced irrelevant features in high-dimensional biomedical datasets.

Conclusions:

• The improved shuffled frog leaping algorithm offers a superior approach for feature selection in high-dimensional biomedical data.
• This method enhances diagnostic accuracy and computational efficiency for molecular diagnosis.
• The proposed enhancements provide a robust solution for complex biomedical data analysis.