AI-Based Automated Feature Selection for Predictive Modeling

Abstract

Artificial Intelligence (AI) has improved predictive modeling by enabling efficient analysis of large-scale and high-dimensional datasets. Before 2019, selecting relevant features from complex data was a major challenge. Feature selection is important for improving accuracy, reducing computational cost, preventing overfitting, and enhancing model interpretability. Traditional methods such as filter, wrapper, and embedded approaches had limitations with big data and nonlinear relationships. To address these issues, AI-based Automated Feature Selection (AFS) techniques were introduced. These methods combine machine learning, optimization algorithms, and intelligent search strategies to automatically identify the most relevant features. Common approaches included Genetic Algorithms (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Artificial Neural Networks (ANN), Support Vector Machines (SVM), Random Forests (RF), and Reinforcement Learning (RL). This study reviews AI-driven feature selection techniques developed between 2009 and 2018 and proposes a hybrid framework combining filter-based preprocessing, evolutionary optimization, and machine learning evaluation. Experimental results show that AI-based hybrid methods outperform traditional techniques in accuracy, scalability, and feature optimization, demonstrating their importance in modern predictive analytics systems.