Deep Learning-Based Severe Dengue Prognosis Using Human Genome Data with Novel Feature Selection Method

Abstract

In recent years, patients affected by dengue fever are getting increased. Outbreaks of dengue fever can be prevented by taking measures in the initial stages. In countries with high disease incidence, it requires early diagnosis of dengue. In order to develop a model for predicting, outbreak mechanism should be clarified and appropriate precautions must be taken. Increase in temperature, sea surface temperature, rapid urbanization and increase of rainfall due to global warming is the interplay factors which influences the outbreaks. Human travel and displacement due to increase in urbanization and temperature, causes dengue virus-infected mosquitoes to be spread. High accurate classification can be achieved by deep learning methods. It is a versatile and regressive method. Small amount of tuning is required by this and highly interpretable outputs are produced. Healthy subjects and dengue patients are differentiated by the factors determined by this and they are used to visualize them also. These factors increase the stability and accuracy of the boosting process in construction of dengue disease survivability prediction model. Problems in overfitting can also be reduced. Applications like decision support systems in healthcare, tailored health communication and risk management are incorporated with the proposed method.