A comparison of normalization techniques in predicting dengue outbreak

In Malaysia, dengue fever (DF) and the potentially fatal dengue hemorrhagic fever (DHF) remain to be a significant public health concern. Higher rainfall and unconcern attitude in the community were some of the factors that contribute to the increase of dengue cases. As number of dengue cases is increasing rapidly in Malaysia, more work need to be done in order to prevent this situation become critical. This includes work on predicting future dengue outbreak. This paper investigates the use of three normalization techniques in predicting dengue outbreak; Min- Max, Z-Score and Decimal Point Normalization. These techniques are incorporated in the LS-SVM and Neural Network (NNM) prediction model respectively. Comparisons of results are made based on prediction accuracy and mean squared error (MSE). Results obtained indicate that the LSSVM is a better prediction model as compared to the NNM.