Weighted Extreme Learning Machine for Dengue Detection with Class-Imbalance Classification

Abstract

Dengue is a disease caused by mosquitoes that may even be lethal to some patients. It is important to detect this disease as soon as possible to decrease the death toll. In this research, we use machines to classify patients as Dengue patients and Non-Dengue patients. The dataset is the treatment data from the patients with fever, cold, flu, pneumonia, and Dengue, from Sarapee Hospital, Chiangmai province, Thailand, during September 2015 to September 2017. The dataset includes 248 records of Dengue patients and 4,960 records of Non-Dengue patients including patient with fever, cold, flu, and pneumonia. We use the text of symptoms of the patients for input data. Weighted Extreme Learning Machine (WELM) is used to solve the class imbalance problems. It was compared for accuracy with neural network and Extreme Learning Machine (ELM). The result shows, that if the number of records of Non-Dengue patients are increasing, the accuracy of the neural network and ELM are decreasing, but the accuracy of WELM is stable.