An improved Differential evolution with Sailfish optimizer (DESFO) for handling feature selection problem

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Mining

Background:

• Feature selection is crucial for machine learning and data mining, especially with high-dimensional data.
• The curse of dimensionality and local optima pose challenges for traditional feature selection algorithms.
• Meta-heuristic techniques offer a promising solution for complex feature selection problems.

Purpose of the Study:

• To introduce a novel hybrid meta-heuristic algorithm, Differential Evolution and Sailfish Optimizer (DESFO), for effective feature selection.
• To evaluate the performance of DESFO against established and modern optimization algorithms.
• To demonstrate the capability of DESFO in improving classification accuracy and reducing feature dimensionality.

Main Methods:

• The proposed DESFO algorithm combines Differential Evolution and Sailfish Optimizer.
• A comparative analysis was conducted against nine other modern algorithms.
• Performance was evaluated using Random Forest and Key Nearest Neighbors classifiers on 14 multi-scale benchmarks.

Main Results:

• DESFO achieved superior performance compared to all other tested algorithms.
• The algorithm significantly improved classification accuracy, reaching 85.7% with Random Forest and 100% with Key Nearest Neighbors.
• Fitness values indicated strong performance, with 71% for Random Forest and 85.7% for Key Nearest Neighbors.

Conclusions:

• The DESFO algorithm is highly effective for feature selection in high-dimensional datasets.
• DESFO offers a robust solution to overcome the limitations of local optima in optimization.
• The proposed method demonstrates significant potential for enhancing machine learning model performance.