A Novel Method Based on Nonlinear Binary Grasshopper Whale Optimization Algorithm for Feature Selection.

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Mining

Background:

• Feature Selection (FS) is crucial for data mining, particularly with high-dimensional data.
• Existing optimization algorithms for FS often lack balanced search capabilities.
• Redundant or irrelevant features can hinder model performance and increase complexity.

Purpose of the Study:

• To propose a novel hybrid optimization algorithm for effective Feature Selection.
• To address the limitations of unbalanced search in current FS optimization techniques.
• To improve the efficiency and accuracy of feature selection in high-dimensional and medical datasets.

Main Methods:

• Developed a nonlinear binary grasshopper whale optimization algorithm (NL-BGWOA).
• Introduced a hybrid position updating strategy combining whale and grasshopper population dynamics.
• Optimized search diversity within the target domain for enhanced feature selection.

Main Results:

• NL-BGWOA demonstrated strong performance on high-dimensional UCI datasets, achieving accuracy up to 0.9895.
• Experimental validation on medical datasets (Parkinson's, COVID-19) showed superior results in accuracy (0.913), feature subset size (5.7), and fitness (0.0873).
• The proposed method proved effective in solving real-world FS problems.

Conclusions:

• The NL-BGWOA offers a comprehensive and superior approach to Feature Selection for high-dimensional data.
• The hybrid optimization strategy effectively balances search diversity and exploitation.
• This algorithm presents significant advantages for applications requiring efficient and accurate feature reduction.