Classification of Lung Cancer Using SVM with Feature Selection Based on PSO-ROC

Abstract

The global issue of lung cancer has grown to be very serious. Using machine learning to classify lung cancer is one method. The challenges in this study are how to apply Particle Swarm Optimization rate of change (PSO-ROC) as a feature selection method and support vector machine (SVM) as a classifier in the context of lung cancer classification; how to compare the accuracy values and running times between SVM without first reducing or selecting the features, SVM with PSO feature selection, and SVM with SVM with PSO-ROC feature selection in the context of lung cancer classification. The purpose of this work is to use SVM with feature selection based on the PSO-ROC algorithm to classify lung cancer. Three methods of classification were used in this study: first, Support Vector Machine (SVM) classification without feature reduction or feature selection; second, SVM and PSO feature selection method; and third, SVM and PSO -ROC feature selection. There are two categories for cancer: malignant and non-cancerous. The findings of this study should help the medical community categorize cancer more quickly and accurately, especially lung cancer. The PSO-ROC based feature selection selects limited number of attributes and yields high classification accuracy compare to others.