Lung Cancer Classification using Fuzzy Interactive Naïve Bayesian Network

Abstract

Lung cancer is the most serious disease in the world and millions of people die of it every year. Because of the limitations of current treatment processes, it is difficult to cure lung cancer if the patient is no longer in the early stages. Therefore, it is necessary to diagnose lung cancer as early as possible, thereby increasing the chances to cure it. The Fuzzy Interactive Naive Bayesian (FINB) network is a new Bayes network that can be used to classify lung cancer by using microarray data sets. The FINB network is an interactive network and every attribution has an interactive parent and with a weight on the relationship that shows the interaction of the attribution in the data set. In our experiments, we use the gene expression profiles from the Affymetrix Human Genome U133 Plus 2.0 microarray. We use the Neural Network with a Weighted Fuzzy Membership Function (NEWFM) to train the data set and reconstruct the Fuzzy Interactive Naive Bayesian network. Then we compare the results with Tree augment naive Bayesian (TAN) network. We conclude that the FINB network performs better than the TAN network.