A Comparative Study on Machine Learning Algorithms and A Hybrid Model of Genetic Algorithm and Neural Network for Mesothelioma Diagnosis

Abstract

Mesothelioma is an aggressive type of lung cancer that is caused by breathing in asbestos fibers. This paper presents an empirical comparative study on the application of various supervised learning algorithms, including logistic regression, random forests, gradient boosting, support vector machine, K- nearest neighbor, Gaussian naive Bayes, and artificial neural network, along with various feature selection methods for Mesothelioma diagnosis. A hybrid approach of genetic algorithm and artificial neural network to data mining on the mesothelioma dataset is also examined. The objective of this study is to find an effective strategy through extensive experimentation on predictive modeling resulting in a classifier of high predictive power for predicting mesothelioma diagnosis. The experimental results show that the hybrid model of genetic algorithm and artificial neural network outperformed other models and achieved AUROC and F1 scores of 0.98 and 0.87 respectively.