Abstract

Deep learning has significantly advanced medical diagnostics through big data utilization; however, high computational costs and "black-box" opacity remain critical barriers in resource-constrained clinical settings. This research investigates randomized machine learning frameworks specifically Extreme Learning Machines (ELMs) and Random Vector Functional Link (RVFL) networks as efficient alternatives. By incorporating stochasticity into the training phase, these architectures reduce computational complexity and latency without compromising diagnostic precision. To mitigate interpretability concerns, the study integrates Explainable AI (XAI) tools, including Local Interpretable Model-agnostic Explanations (LIME) and Shapley Additive Explanations (SHAP), to elucidate model decision-making. Empirical evaluations involving genitourinary cancers and coronary artery disease datasets reveal that RVFL superiorly balances performance and efficiency, achieving accuracies of 88.29% and 81.64% with significantly reduced processing times (6.22s and 0.0308s, respectively). These findings advocate for the adoption of randomized models to enhance transparency and operational speed in healthcare, ultimately facilitating more accessible, interpretable, and timely diagnostic interventions.