Machine learning classifiers for predicting the presence of cancer using gene expression data from CTCs/CTMs

Abstract

This thesis focuses on developing and evaluating machine learning classifiers for predicting the presence of seven types of cancer using gene expression data from CTCs and CTMs. The cancers investigated include liver cancer, breast cancer, colorectal cancer, non small cell lung cancer, pancreatic cancer, prostate cancer, and melanoma. The study involves building binary classifiers to distinguish each cancer type from others and multiclass classifiers to predict all seven cancer types. The goal is to compare these approaches and identify the most effective model for accurate cancer prediction. The findings demonstrate the significant potential of machine learning models in enhancing cancer diagnostics using minimally invasive methods.Among the models evaluated, the Random Forest multi-classifier emerged as the most reliable and effective, making it highly recommended for practical use in cancer diagnosis.