A Prediction Modelling for Dengue Fever Outbreaks Using Data Mining Techniques

Abstract

Dengue fever has become one of Thailand's most critical concerns in recent years. The number of dengue cases from January to May 2021 in the area under the responsibility of the Office of Disease Prevention and Control 2 in Phitsanulok was the second highest in Thailand. Therefore, the objective of this research was to construct a reasonable model for predicting dengue fever outbreaks in Phitsanulok using data mining techniques. The data used were the dengue fever patient information from the Office of Disease Prevention and Control 2 in Phitsanulok, and the monthly weather data from the Northern Meteorological Center of 5 provinces. Four parameters were used to predict or categorize the symptoms of Dengue fever including personal data, time, location, and climate from January 1, 2009, to May 31, 2021, with a total of 47,386 items and 12 attributes. The data was evaluated by using four different data mining techniques which were Decision trees, Naïve Bayes, Artificial neural network (ANN), and Support vector machines techniques (SVM). Four attributes of the district, province, year, and monthly average precipitation were chosen to construct the models by using the selection technique. The results revealed that the most suitable model for forecasting dengue fever outbreaks was obtained from the Decision trees technique. It gave the highest accuracy and the f-measure of 69.83% and 75.4%, respectively. The forecasting model developed in this study could be applied for policy planning and campaigning to prevent the dengue epidemic in the high-risk population under limited budgets and resources.