Dengue Prediction using Machine Learning Algorithms

Abstract

Dengue is an arboviral disease caused by the Aedes mosquito-borne dengue viruses (DENVs). The World Health organization (WHO) reports an annual incidence of around 100-400 million infections in 2019 which is the largest number of dengue cases ever reported globally and prompted WHO to declare the virus as the world's top 10 public health threats. It can be turned into life-threatening dengue hemorrhagic fever which further evolves into dengue shock syndrome. Indispensable useful tools that precisely distinguish dengue and its subtypes in the early stage of disease progression are essential to convenient well-timed supportive care and therapy. In recent years, Bangladesh has seen a hike in the dengue outbreak and 101,000 cases were reported to the WHO in 2019. Such an outbreak can create havoc in society. Because of the lack of vaccine and antiviral drugs, a timely prediction of the dengue outbreak is, therefore, crucial to reducing the casualty. In this paper, we proposed a new machine learning approach to predict dengue fever. A patient dataset, containing information of the patient's diagnosis report, medical history, and symptoms, was constructed through collecting real-time raw data samples of various types of dengue fever patients from the Medicine Department of Chittagong Medical College Hospital and Dhaka Medical College Hospital, Bangladesh. The whole dataset was split into 70:30 ratios using 70% for training and 30% for test purposes. We applied machine learning algorithms, namely decision tree (DT) and random forest (RF) in the proposed classification model. Finally, the decision tree resulted in an average accuracy of 79%, which is higher than the random forest.

Topics & Concepts