Ensemble Feed-Forward Neural Network and Support Vector Machine for Prediction of Multiclass Malaria Infection

Globally, recent research are focused on developing appropriate and robust algorithms to provide a healthcare system that is versatile accurate. Existing malaria models plagued with low rate of convergence, overfitting, limited generalization due restriction binary cases prediction, proneness local minimum errors in finding reliable testing output complexity features the feature space, which black box nature. This study adopted stacking method heterogeneous ensemble learning ArtificialNeural Network (ANN) Support Vector Machine (SVM) predict multiclass, symptomatic, climatic infection. ANN produced 48.33 percent accuracy, 60.61 sensitivity, 45.58 specificity. SVM Gaussian kernel function gave better performance results 85.60 84.06 86.09 Consequently, improve prediction performance, was introduced ANN. The proposed model tuned different thresholds at threshold value 0.60, an optimum accuracy 99.86 percent, sensitivity 100 specificity 98.68 mean square error 0.14. experimental indicated stacked multiple classifiers than single model. demonstrated efficiency effects variations multiclass infection classification. Furthermore, reduced complexity, problems comparison previous related models.