Hybrid Binary Dragonfly Algorithm with Simulated Annealing for Feature Selection.

Area of Science:

• Computer Science
• Machine Learning
• Data Science

Background:

• High dimensionality in data increases memory and computational demands.
• It also negatively impacts machine learning classifier performance.
• Feature selection is crucial for mitigating these issues.

Purpose of the Study:

• To address the challenges posed by high-dimensional data.
• To improve the feature selection process for classification tasks.
• To enhance the accuracy and efficiency of machine learning models.

Main Methods:

• An improved Binary Dragonfly Algorithm (BDA) was developed by integrating Simulated Annealing (SA).
• The hybrid algorithm, named BDA-SA, aims to overcome local optima problems in the standard DA.
• SA is applied to refine the best solutions from the BDA for better feature subset selection.

Main Results:

• The BDA-SA approach was evaluated using benchmark datasets from the UCI repository.
• The proposed hybrid feature selection (FS) method demonstrated superior performance.
• BDA-SA outperformed the basic Binary Dragonfly Algorithm and other wrapper-based FS methods.

Conclusions:

• The BDA-SA algorithm effectively improves feature selection for classification.
• This hybrid approach enhances machine learning classifier accuracy and efficiency.
• BDA-SA offers a robust solution for managing high-dimensional data challenges.