Feature importance feedback with Deep Q process in ensemble-based metaheuristic feature selection algorithms.

Area of Science:

• Machine Learning
• Data Science
• Artificial Intelligence

Background:

• Feature selection is crucial for high-dimensional data to prevent overfitting and improve efficiency.
• Existing methods like filter, wrapper, and embedded approaches have limitations in robustness and computational load.
• Metaheuristic algorithms (PSO, FA, WOA) are useful but struggle with incorporating feature importance feedback.

Purpose of the Study:

• To develop a novel ensemble feature selection model.
• To enhance metaheuristic algorithms with Deep Q-Learning for iterative relevance feedback.
• To improve the effectiveness and efficiency of feature selection in machine learning.

Main Methods:

• An ensemble model integrating Particle Swarm Optimization (PSO), Firefly Algorithm (FA), and Whale Optimization (WOA).
• Incorporation of a Deep Q-Learning framework for intelligent feature importance updates based on model performance.
• Iterative fine-tuning of the feature selection process through relevance feedback.

Main Results:

• The proposed ensemble model achieved significant performance gains over traditional and individual metaheuristic methods.
• Demonstrated improvements include 9.5% higher precision, 8.5% higher accuracy, 8.3% higher recall, 4.9% higher AUC, and 5.9% higher specificity.
• Validation across multiple software bug prediction datasets and samples confirmed the model's effectiveness.

Conclusions:

• The novel feature selection framework effectively addresses limitations of existing methods.
• The approach offers superior performance metrics and a flexible architecture for diverse machine learning applications.
• This research paves the way for more robust and efficient machine learning models in fields like healthcare and NLP.