An Explainable AI Based Approach for Forecasting Dengue Diseases Using Patient’s Phenotypic Traits

Dengue fever, a deadly viral infection, is a growing concern globally, particularly in Bangladesh. The disease’s impact is increasing daily, and early detection and intervention are crucial. A study was conducted to explore the predictive capabilities of machine learning in identifying dengue fever symptoms and assessing infection likelihood. The study used real-time data from Khulna Medical College Hospital and Ad-din Akij Medical College Hospital to compile a patient dataset of 721 examples with 20 attributes. Four machine learning algorithms were used: logistic regression, random forest, extreme gradient boosting, and artificial neural networks. The Random Forest model emerged as the most effective, with an impressive accuracy rate of 84.14%. The precision and sensitivity metrics also emphasized its reliability. The study also conducted a SHAP analysis, revealing that the most significant factors were fever, gender, age, pulse, and platelet count. The findings highlight the potential of machine learning in dengue fever prediction and the importance of leveraging such technologies in public health initiatives.