Every single minute, enormous amounts of data is produced, due to the rapidadvancement and extensive use of technology. However, managing andanalyzing this data has become more difficult, particularly when there are morefeatures than observations in high-dimensional situations. Classifier performancein supervised machine learning problems may significantly be affected by thishigh dimensionality. Ensemble learning, a popular machine learning technique,emerged as an effective solution for this particular issue. ENet, SVM, NN are thethree base learners that are used in this proposed ensemble framework. Thisensemble reduces the misclassification error and also enhancing robustness ascompared to other classifier. The proposed method selects predictions based onmaximum True Positive Rate (TPR) of the base classifier. In order to aggregatethe predictions of the various classifiers, majority voting is used. We analyzedour ensemble's performance (called ETP) on simulated datasets and microarraydatasets. The ETP constantly achieves minimum misclassification error rates onsimulated datasets and microarray datasets which indicate that it perform betteras compared to other well-known classifiers, such as RF, KNN, AdaBoost andthe baseline classifiers that were used to build it.