Machine Learning Applications for Solving Complex Optimisation Problems Across Various Domains

Abstract

Optimization of complex problems is fundamental across industries, yet traditional methods often struggle with high-dimensionality, nonlinearity, and dynamic constraints. This study aims to explore how machine learning (ML) techniques address these challenges by enhancing optimization across diverse sectors, including engineering, healthcare, energy systems, finance, and transportation. A systematic review and analysis were conducted by mapping ML methods to specific domain applications, assessing their distribution and impact. Results show that reinforcement learning dominates with a 25% share across applications, followed by supervised and deep learning techniques, each representing 16.7% of total usage. Specialized approaches such as convolutional neural networks, predictive modeling, anomaly detection, Bayesian optimization, and sensor fusion account collectively for 41.6%, reflecting the growing diversity of ML-driven solutions. The novelty of this work lies in its cross-domain integration, quantifying how ML methods not only replace but enhance traditional optimization approaches through adaptability, scalability, and real-time decision-making capabilities. Additionally, all examined domains exhibited three major ML application areas, indicating a uniform breadth of adoption. In conclusion, ML is redefining optimization practices, offering dynamic, intelligent, and domain-adaptive solutions. Future directions are suggested toward enhancing interpretability, robustness under uncertainty, and cross-domain generalization. This study was fully self-funded through contributions from all authors without external financial support.