IBGJO: Improved Binary Golden Jackal Optimization with Chaotic Tent Map and Cosine Similarity for Feature Selection

Area of Science:

• Machine Learning
• Data Mining
• Optimization Algorithms

Background:

• Feature selection is vital for improving predictive model efficacy and precision.
• Reducing features enhances classification accuracy and lessens computational load.
• Conventional optimization algorithms can suffer from premature convergence and limited diversity.

Purpose of the Study:

• To propose an improved binary golden jackal optimization (IBGJO) algorithm for wrapper-based feature selection.
• To enhance the convergence rate and accuracy of the golden jackal optimization (GJO) algorithm.
• To evaluate the effectiveness of novel mechanisms for feature selection.

Main Methods:

• Developed the improved binary golden jackal optimization (IBGJO) algorithm.
• Incorporated a chaotic tent map (CTM) for population initialization and diversity.
• Implemented an adaptive position update mechanism using cosine similarity to prevent premature convergence.
• Utilized a binary mechanism tailored for binary feature selection problems.

Main Results:

• IBGJO demonstrated significantly improved convergence rate and accuracy compared to conventional GJO and other algorithms.
• The CTM mechanism and cosine similarity-based position update enhanced exploitation and population diversity.
• Empirical results on 28 UCI datasets validated the effectiveness of the proposed enhancements.
• The IBGJO algorithm showed faster convergence and superior performance in feature selection tasks.

Conclusions:

• The proposed IBGJO algorithm offers a robust and efficient approach to feature selection.
• The CTM mechanism and cosine similarity strategy effectively address limitations of the conventional GJO algorithm.
• IBGJO provides a valuable tool for enhancing machine learning model performance through optimized feature selection.