AVALIAÇÃO DE TÉCNICAS DE APRENDIZADO DE MÁQUINA NA CLASSIFICAÇÃO DO CÂNCER

Abstract

The present work was based on the importance of early cancer detection. A wide variety of microRNAs (miRNAs) that specifically indicate many types of cancer can be identified and their miRNA expression profiles analyzed. Thus, miRNAs serve as a biopsy diagnostic tool non-invasive liquid for the early detection of many types of cancer. Therefore, it was possible to evaluate machine learning techniques in cancer classification, based on a dataset obtained from the Albert Einstein Hospital. Three classification models were tried, two Convolutional Neural Networks and one Support Vector Machine. As a result, it was possible to train a Convolutional Neural Network with structured data, but such a set needs to be balanced. And if the dataset has only two classes, classification models with Support Vector Machine show the best performance in classification.