Comparative Study on Data Mining Techniques Applied to Breast Cancer Gene Expression Profiles

摘要

Breast cancer has the second highest incidence among all cancer types and is the fifth cause of cancer related death among women. In Brazil, breast cancer mortality rates have been rising. Cancer classification is intricate, mainly when differentiating subtypes. In this context, data mining becomes a fundamental tool to analyze genotypic data, improving diagnostics, treatment and patient care. As the data dimensionality is problematic, methods to reduce it must be applied. Hence, the present study aims at the analysis of two data mining methods (i.e., decision trees and artificial neural networks). Weka and MATLAB were used to implement these two methodologies. Decision trees appointed important genes for the classification. Optimal artificial neural network architecture consists of two layers, one with 99 neurons and the other with 5. Both data mining techniques were able to classify data with high accuracy.