Enhanced feature selection technique using slime mould algorithm: a case study on chemical data.

Area of Science:

• Computational intelligence
• Data science
• Bioinformatics

Background:

• Feature selection (FS) is crucial for data analysis and decision-making, impacting classification accuracy and efficiency.
• Metaheuristic (MH) algorithms offer promising approaches for optimizing FS processes.
• Existing FS methods can struggle with convergence and local optima, necessitating improved techniques.

Purpose of the Study:

• To develop a novel hybrid metaheuristic algorithm, SMAMPA, for enhanced feature selection.
• To improve convergence rates and avoid local optima in feature selection.
• To evaluate SMAMPA's performance on diverse datasets and its applicability to real-world problems like Quantitative Structure-Activity Relationship (QSAR) modeling.

Main Methods:

• A modified Slime Mould Algorithm incorporating Marine Predators Algorithm operators for local search.
• Evaluation of the SMAMPA algorithm on twenty benchmark datasets.
• Comparison of SMAMPA against state-of-the-art feature selection methods.
• Application of SMAMPA for Quantitative Structure-Activity Relationship (QSAR) modeling.

Main Results:

• SMAMPA demonstrated a significant increase in prediction rates across tested datasets.
• The method effectively reduced dataset dimensionality.
• SMAMPA outperformed existing feature selection techniques in various performance metrics.
• Successful application of SMAMPA in QSAR modeling indicated its real-world utility.

Conclusions:

• The developed SMAMPA algorithm is highly effective for feature selection, improving prediction accuracy and reducing data dimensions.
• SMAMPA offers a robust and efficient solution for complex data analysis tasks.
• The hybrid approach successfully addresses limitations of traditional metaheuristic feature selection methods.