A Novel Feature Selection Method Based on Extreme Learning Machine and Fractional-Order Darwinian PSO.

Area of Science:

• Machine Learning
• Computational Intelligence
• Data Science

Background:

• Feature selection is crucial for improving regression model performance and reducing complexity.
• Extreme Learning Machines (ELM) offer efficient training but require effective feature selection.
• Particle Swarm Optimization (PSO) is a metaheuristic optimization algorithm widely used in feature selection.

Purpose of the Study:

• To propose a novel feature selection method combining Extreme Learning Machine (ELM) and Fractional-order Darwinian Particle Swarm Optimization (FODPSO) for regression problems.
• To enhance the performance of regression models by identifying optimal feature subsets.
• To evaluate the proposed method's effectiveness compared to existing approaches.

Main Methods:

• The proposed method utilizes ELM to construct a fitness function based on Mean Square Error (MSE).
• An improved PSO algorithm, FODPSO, is employed to search for the optimal solution of the fitness function.
• Comparative experiments were conducted on seven public datasets to assess the method's performance.

Main Results:

• The proposed FODPSO-ELM method achieved the lowest MSE values in six out of seven comparative experiments.
• The method demonstrated superiority in achieving lower MSE with the same feature subset size.
• It also showed the ability to find a smaller feature subset for comparable MSE values.

Conclusions:

• The FODPSO-ELM approach is a highly effective and superior method for feature selection in regression tasks.
• The proposed technique offers improved accuracy and efficiency in identifying relevant features.
• This method provides a valuable contribution to the field of machine learning and data analysis.