Evaluating Performance of Regression and Classification Models Using Known Lung Carcinomas Prognostic Markers

Abstract

Differential expression study between tumor and non-tumor cells aids lung cancer diagnostic classifications and prognostic prediction at various stages. Support vector machine (SVM) learning is used to categorize the morphology of lung cancer. Logistic regression, random forest, and group lasso-based models are used to model dichotomous outcome variables. The purpose is to take groups of observations and design boundaries to forecast which group future observations belong to base measurements. The performance of these selected regression and classification models using lung cancer prognostic indicators is evaluated in this article. The presented results might guide for further regularizations in classification techniques using known lung carcinoma marker genes.