Migrating birds optimization-based feature selection for text classification.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Text classification faces challenges with high-dimensional feature sets.
• Existing feature selection methods often lack efficiency and scalability.
• Suboptimal accuracy and high computational overhead are common issues.

Purpose of the Study:

• To propose a novel methodology, MBO-NB, for effective feature selection in text classification.
• To enhance both computational efficiency and classification accuracy.
• To address limitations of current techniques in handling extensive feature sets.

Main Methods:

• Integration of Migrating Birds Optimization (MBO) with Naïve Bayes classifier.
• Preprocessing using the Information Gain algorithm for feature reduction.
• Comparative analysis against existing techniques, including Particle Swarm Optimization (PSO).

Main Results:

• MBO-NB significantly reduces feature count from an average of 62,221 to 2,089.
• Demonstrated superior effectiveness in feature reduction and classification accuracy.
• Outperformed PSO by an average of 6.9% across four experimental setups.

Conclusions:

• MBO-NB offers a scalable and effective solution for text classification feature selection.
• The integration of MBO and Naïve Bayes provides a comprehensive approach.
• Contributes to advancing text classification techniques for more robust systems.