Data mining in Breast Cancer

Abstract

Machine learning and data mining methods can be the future of the clinical decision process like pathological diagnosis. In this project we studied Breast Cancer Wisconsin dataset and applied different algorithms, concretely classifiers, in order to predict the diagnosis and the prognostic of the cancer. In order to classify the different types of cancer we divided the classification in two steps and we tested different algorithms for each step. The first step is the diagnosis classification. Diagnosis consistsin predict if the cancer is malignant and benign. And the second step is the prognostic classification. Prognostic consist in predict if cancer is recurrent or non-recurrent. After applying different models for each steps the result is that the best model to predict the diagnosis is the Decision Forest model. And the best model to predict the prognostic is the Boosted Decision Tree model. So, we conclude that the two step classifier with Decision Forest model and Boosted Decision Tree model is the best classifier.