Combating Breast Cancer by an Intelligent Ensemble Classifier Approach
Tawseef Ayoub Shaikh, Rashid Ali · 2018
Breast cancer is the most vulnerable type of cancers and has posed a serious threat and challenge to human existence. Every domain of human innovation is trying in tackling this threat and the present IT age has also contributed for this noble cause through the innovative innovation of machine learning concepts. Here in this work, we present an intelligent ensemble classifier mix up using the all-time WEKA data mining toolkit on BCDR-F03 benchmark cancer dataset. Five individual classifiers Naive Bayes, SVM, Simple Logistics, Random Forest and iBK (Instance Based Learning) from different groups in machine learning are evaluated on selected dataset and their corresponding individual accuracies came out to be 76%, 80%, 80%, 81 % and 79. 89% respectively. The final results revealed that three ensembles picked up from voting i.e. Voting Ensemble (SMO, RF, iBK), Voting Ensemble (SL, RF, iBK) and Voting Ensemble (SGD, SL, iBK) got highest accuracies of 83.0163%, 83.1522% and 82.7446% respectively. Similarly in stacking, Stacking Ensemble (SMO, RF, iBK, NB), Stacking Ensemble (SMO, RF, iBK) and Stacking Ensemble (SMO, RF) got highest accuracies of 82.7446%, 83.0163% and 81.9293%, when the combining algorithm used is Simple Logistics, respectively. The results clearly are far better than the individual performance of ML algorithms on our dataset.