An Electric Fish-Based Arithmetic Optimization Algorithm for Feature Selection.

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Science

Background:

• Intelligent information systems generate massive datasets with irrelevant, noisy, and redundant features, necessitating efficient feature selection (FS).
• Swarm intelligence and artificial methods have emerged as powerful tools for addressing complex problems, including FS in high-dimensional data.

Purpose of the Study:

• To propose an innovative hybrid feature selection approach, EFOAOA, by integrating Electric Fish Optimization (EFO) and Arithmetic Optimization Algorithm (AOA).
• To enhance the exploration capabilities of EFO for high-dimensional feature selection problems, aiming for improved convergence speed and accuracy.

Main Methods:

• A hybrid algorithm, EFOAOA, was developed by combining EFO and AOA to leverage their respective strengths in optimization.
• The proposed EFOAOA was evaluated on eighteen diverse real-life datasets.
• Performance was benchmarked against state-of-the-art optimizers using statistical metrics and the Friedman test.

Main Results:

• The integration of AOA significantly boosted EFO's performance in feature selection.
• EFOAOA demonstrated high accuracy and efficiency in identifying the most relevant features.
• The proposed method achieved the lowest number of features and highest accuracy on 50% and 67% of the tested datasets, respectively.

Conclusions:

• The hybrid EFOAOA approach is effective for high-dimensional feature selection.
• The proposed method offers a promising solution for managing complex datasets in intelligent information systems.
• EFOAOA shows superior performance compared to existing feature selection techniques.