Computational Model for Prediction of Malignant Mesothelioma Diagnosis.

This article focuses on applying and comparing various machine learning techniques to improve the diagnosis of malignant pleural mesothelioma (MPM), an aggressive lung cancer linked to asbestos exposure, using a publicly available dataset from the University of California Irvine (UCI) machine learning repository. The study addresses key challenges such as class imbalance by employing oversampling methods including resampling, synthetic minority oversampling technique (SMOTE), and adaptive synthetic sampling (ADASYN), and applies multiple feature selection methods like principal component analysis (PCA), ordinary least squares (OLS), random forest feature selection (RFFS), and genetic algorithms (GA) to identify relevant predictors. Results indicate that artificial neural networks (ANN) combined with the resampling strategy and PCA-selected features achieved the highest accuracy of 96%, outperforming previous studies that often used a duplicated target feature and did not handle class imbalance properly. Ensemble methods, particularly stacking classifiers, also showed strong performance, especially with SMOTE and ADASYN balanced data. The study concludes that addressing data imbalance and careful feature selection are critical for reliable MPM diagnosis models and suggests further exploration of these approaches in other medical contexts.