A Novel Adaptive Sand Cat Swarm Optimization Algorithm for Feature Selection and Global Optimization.

Area of Science:

• Machine Learning
• Data Mining
• Optimization Algorithms

Background:

• Feature selection is crucial for machine learning and data mining.
• Selecting optimal features from high-dimensional datasets presents a significant challenge.
• Existing algorithms may struggle with global search capacity and convergence rates.

Purpose of the Study:

• To enhance the feature selection process using an improved optimization algorithm.
• To augment the global search capacity and convergence rate of the Sand Cat Swarm Optimization algorithm.
• To address the challenge of optimal feature selection in high-dimensional datasets.

Main Methods:

• Developed an enhanced Sand Cat Swarm Optimization algorithm (MSCSO).
• Incorporated logistic chaotic mapping and lens imaging reverse learning for population initialization.
• Utilized nonlinear parameter processing for balancing exploration and development.
• Implemented Weibull flight, triangular parade, and Gaussian-Cauchy mutation strategies for position updates and local optima avoidance.

Main Results:

• MSCSO demonstrated strong performance on 65.2% of CEC2005 benchmark test functions.
• Achieved the best average fitness on 93.3% of UCI datasets.
• Reduced feature selection by 86.7% while maintaining 100% best average accuracy across datasets.
• Significantly outperformed comparative algorithms in feature selection tasks.

Conclusions:

• The enhanced MSCSO algorithm offers superior performance in feature selection.
• MSCSO effectively balances global exploration and local exploitation for optimization.
• This method significantly enhances model accuracy and efficiency in machine learning applications.