Feature Selection with Fast Correlation-Based Filter for Breast Cancer Prediction and Classification Using Machine Learning Algorithms

Abstract

Breast cancer is the first female cancer responsible for high mortality worldwide. Despite the progress that has made it possible to better understand the mechanisms of cancer development, the causes of breast cancer are currently unknown. Nevertheless, studies have identified some risk factors that promote breast cancer and a healthy lifestyle can reduce risk. In Morocco, breast cancer is the first cancer in women. It represents 34.3% of all female cancers. In this work, the Fast Correlation-Based Feature selection (FCBF) method is used to filter irrelevant and redundant characteristics in order to improve the quality of cancer classification, and we will provide an overview of the evolution of key data in the health system and apply five learning algorithms to a breast cancer data set. The purpose of this research work is to predict breast cancer, using several machine-learning algorithms that are Random Forest, Naive Bayes, Support Vector Machines SVM, K-Nearest Neighbors K-NN, and Multilayer Perception MLP, in order to select the most effective algorithm with and without FCBF. The experimental results show that SVM gives the highest accuracy of 97.9% without FCBF but if we apply this method we find that the SVM and MLP show the best results in comparison with other algorithms. The results will help to choose the best learning algorithm machine classification for breast cancer prediction.