FTDZOA: An Efficient and Robust FS Method with Multi-Strategy Assistance.

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Mining

Background:

• Feature selection (FS) is crucial for big data analytics, reducing redundancy and optimizing resources.
• Traditional Zebra Optimization Algorithm (ZOA) faces challenges like local optima and slow convergence in FS.

Purpose of the Study:

• Introduce an enhanced Zebra Optimization Algorithm (ZOA), named FTDZOA, for superior feature dimensionality reduction.
• Address limitations of ZOA, including local optima, limited global search, and slow convergence for FS problems.

Main Methods:

• Integrate a fractional order search strategy to enhance ZOA's exploitation capabilities.
• Incorporate a triple mean point guidance strategy to improve ZOA's exploration prowess.
• Introduce a differential strategy to further elevate ZOA's exploration capacity by integrating information disparities.

Main Results:

• FTDZOA achieved higher classification accuracy on over 90% of tested datasets.
• FTDZOA demonstrated a winning rate exceeding 83% in terms of execution time compared to nine advanced FS methods.
• The FTDZOA-based FS method was applied to 23 FS problems across low, medium, and high dimensions.

Conclusions:

• FTDZOA is a reliable, high-performance, practical, and robust feature selection method.
• The enhanced algorithm effectively addresses ZOA's limitations in feature selection tasks.
• FTDZOA offers significant improvements in both accuracy and efficiency for dimensionality reduction.