A Comparative Study of Machine Learning Techniques for Multi-Class Classification of Arboviral Diseases

Abstract

Among the neglected tropical diseases (NTDs), arboviral diseases present a significant number of cases worldwide. Their correct classification is a complex process due to the similarity of symptoms and the lack of tests in Brazil countryside is a big challenge to be overcome. Given this context, this paper proposes a comparative study of machine learning techniques for multi-class classification of arboviral diseases, which considers three classes: DENGUE, CHIKUNGUNYA and OTHERS, and uses clinical and socio-demographic data from patients. Feature selection techniques were also used for selecting the best subset of attributes for each model. Gradient boosting machines presented the best result in the metrics and a good subset of attributes for daily usage by the physicians that resulted in a 76.58% recall on the CHIKUNGUNYA class.