Machine learning-driven haematological feature analysis for early diagnosis and severity stratification of dengue

Dengue fever is a mosquito-borne viral infection that creates a significant public health challenge in India. The disease originated from four closely related viruses transmitted by the Aedes mosquito, primarily Aedes aegypti. It is widespread in tropical and subtropical regions. It necessitates training algorithms to make predictions or decisions based on data. Developing accurate Dengue prediction models based on machine learning is essential for timely intervention and resource allocation. To develop the model, we need to identify the important clinical markers of dengue patients with feature engineering. A model is developed to predict dengue based on haematological features. We utilized the publicly available Figshare dataset derived from Yasuda et al.'s multicenter observational study in Manila (N=606), which profiles time-series CBC parameters, including the novel Immature Platelet Fraction across dengue and bacterial infections. Dengue and non-dengue classification is carried out by using machine learning classifiers, Random Forest, and XGBoost. CRP, Platelet count, and WBC count were the top parameters to predict dengue for these models. Additionally, the severity classification of Dengue patients is performed using Random Forest and XGBoost. The accuracy achieved for Random Forest and XGBoost is 0.93 and 0.96, respectively. More machine learning classifiers and ensemble models will need to be implemented to facilitate further comparative studies.