Systematic Review on Breast Cancer Classification Using Random Forest and Extreme Learning Machine: Cost Sensitivity and Computational Complexity Perspectives

Abstract

Breast cancer remains one of the most common and deadly cancers affecting women worldwide. Early detection and accurate diagnosis are essential to improve patient survival rates and reduce long-term treatment costs. With the advancement of digital technologies, machine learning (ML) has emerged as a powerful tool in breast cancer classification. Among various ML algorithms, Random Forest (RF) and Extreme Learning Machine (ELM) have gained prominence due to their predictive capabilities. This systematic literature review aims to compare the classification performance of RF and ELM, focusing on cost sensitivity and computational complexity. Using PRISMA guidelines, 60 peer-reviewed articles published between 2013 and 2024 were analyzed. The findings show that RF generally offers high accuracy and robustness against overfitting, making it suitable for complex clinical datasets. Conversely, ELM excels in training speed and computational efficiency, making it ideal for real-time diagnostic systems. However, both methods face challenges in handling imbalanced data, where misclassification of malignant cases can be fatal. Cost-sensitive learning strategies are shown to improve model sensitivity toward minority classes, though their integration into ELM remains limited. Furthermore, computational efficiency is a critical factor, particularly in resource-constrained medical environments. This review provides a thematic synthesis of current research and highlights future directions, such as developing hybrid models combining RF’s accuracy with ELM’s efficiency, and implementing explainable AI for trustworthy clinical adoption.