SVM-PSO based Feature Selection for Improving Medical Diagnosis Reliability using Machine Learning Ensembles

Improving accuracy of supervised classification algorithms in biomedical applications,\udespecially CADx, is one of active area of research. This paper proposes construction of rotation\udforest (RF) ensemble using 20 learners over two clinical datasets namely lymphography and\udbackache. We propose a new feature selection strategy based on support vector machines\udoptimized by particle swarm optimization for relevant and minimum feature subset for obtaining\udhigher accuracy of ensembles. We have quantitatively analyzed 20 base learners over two\uddatasets and carried out the experiments with 10 fold cross validation leave-one-out strategy\udand the performance of 20 classifiers are evaluated using performance metrics namely accuracy\ud(acc), kappa value (K), root mean square error (RMSE) and area under receiver operating\udcharacteristics curve (ROC). Base classifiers succeeded 79.96% & 81.71% average accuracies\udfor lymphography & backache datasets respectively. As for RF ensembles, they produced\udaverage accuracies of 83.72% & 85.77% for respective diseases. The paper presents promising\udresults using RF ensembles and provides a new direction towards construction of reliable and\udrobust medical diagnosis systems