LUNG CANCER PREDICTION USING GENE EXPRESSION DATA WITH STATISTICAL APPROACH FOR GENE SELECTION

Abstract

Cancer is caused by abnormal cell growth or cell division in the body. This uncontrolled and undesired growth is a reflection of genetic variation causing abnormal functioning of genes, causing a change in gene expression. This change in gene expression is brought understudy for cancer prediction, diagnosis, and treatment. Machine Learning techniques when applied to gene expressions can predict one’s susceptibility towards cancer. The tough task is to determine those genes that possess a stronger capability or show greater variation in expression when in an abnormal state than the normal state. This paper proposes gene selection techniques for selecting an optimal subset of genes that are highly important for accurate prediction. The lung cancer gene expression data has been taken from Kent Ridge Biomedical Dataset Repository. The main focus of the project is to select the optimal subset of genes to have a high value of AUC_ROC and F-measure to make a correct assessment of the model dealing with an imbalance dataset.