Developing an Effective Predictive Model for Clinical Dataset

The classification falls into the essential tasks of supervised learning. The prediction of highly disperse nature of data, such as clinical medical data, is an open challenge. This study performs experimental analysis of machine learning algorithm to develop an effective prediction model for the clinical dataset. A real time clinical dataset has been considered for the model development. The machine learning algorithms: Decision Tree (J48), Naïve Bayes (NB), Random Forest (RF) and Multilayer Perceptron (MLP) have been applied. The validation of prediction model has been evaluated using both cross validation and split percentage methods. It has been evident that classification algorithms in combination of ensemble methods perform better. Thus, an effective prediction model performs better in presence of ensemble methods. The bagging, voting and stacking - ensemble methods have been used in this study. The experimental outcomes revealed that Naïve Bayes produced better predictive results using 10-fold cross validation method. Whilst, MLP outperformed using split percentage method for evaluation.