Optimized Ensemble Prediction Model for Breast Cancer

Abstract

Breast cancer-associated to females has been reckoned as one of the most prevalent cancers. For better medical treatments premature detection of breast cancer is an essential step. This study focuses on automated breast cancer prediction using the Ensemble Machine learning paradigm. Supervised machine learning models are trained using labelled data to perceive a hypothesis that will give good predictions for a particular problem domain. Although the hypothesis space contains hypotheses that are very well-suited for a particular problem, it may be very difficult to find a good one. Ensemble learning combines multiple learnings to form a better hypothesis. The expression Ensemble is usually reserved for methods that generate predictions from various hypotheses using homogeneous or non-homogeneous base learners. Additional computation is typically required in assessing such types of ensemble models than evaluating the prediction from a single model. Unlike bagging or boosting, we are using non-homogeneous classifiers to predict whether the breast cancer is cancerous or not that is, malignant or benign using GaussianNB as meta classifier in stacking classifier of sci-kit learn in python and we are using breast cancer dataset from Wisconsin, maintained by the University of California. The recorded prediction was achieved to be 99.41% which outperforms the performance of the single algorithm.