The Model Development for Forecasting the NS1Ag Antigen Examination Results of Dengue Fever Patients Using the Multi-Layer Perceptron Neural Network

Abstract

The objective of the article is to develop a model for forecasting the NS1Ag antigen examination results of dengue fever patients by using the Multi-Layer Perceptron Neural Network. The data used for developing the model was the results of the antigen type NS1Ag tests collected from a certain hospital's information system. The data dated from October 2018 -March2020 with the number of 1,916 records. The data featured with 11 attributes.The Resample and SMOTE techniques were applied to adjust the imbalance ofthe data by increasing the size of data from 100% to 200%. The effectivenessevaluation results of the model using the percentage split of 70:30 revealedthat the imbalance-adjusted model using the SMOTE method with the size of200% and the Multi-Layer Perceptron Neural Network technique provided theaccuracy of 93.54%. This provided higher accuracy compared to the modelusing the Multi-Layer Perceptron Neural Network solely. The results of thestudy show that the model could be used for forecasting the NS1Ag antigenexamination results of dengue fever patients. It can be applied as a tool tosupport doctors and nurses to diagnose the dengue fever accurately and filterpatient transfers to the laboratory appropriately.