A Study on Hybrid Ensemble Approaches for Effective Dengue Detection and Classification

Abstract

Dengue is one of the most prevalent mosquito-borne viral diseases and a global public health concern in tropical and subtropical countries. Timely diagnosis of dengue is essential to prevent complications and death. This work introduces an improved hybrid ensemble model named as HyDengue to improve the accuracy and trustiness of dengue classification based on blood data. Then the white blood cells datasets were preprocessed rule based on data cleaning, feature selection and normalization as well as SMOTE-based balancing of missing and unequal number of samples. HyDengue combines several intelligent classifiers as ensemble and also applies stacked ensemble framework that harnesses the learning ability of its base models complementarily to get improved performance. Experimental results show that under the 5- fold cross validation testing, HyDengue achieved a training accuracy of 99.7%, and a testing accuracy of 97.6% with an F1-score of:0.974 and an ROC-AUC of:0.998 better than their counterpart classical machine learning models. Execute evaluation section on behalf to leave space faire: So for sure more challenging pandemic tools. These results reveal that the developed framework have stable learning capability for classification and strong generalization ability without overstated easily. The high precision and recall rates of the model demonstrate its potential to be able to accurately detect dengue-positive cases without having too many false negatives. The study proves that ensemble learning technique is effective for clinical diagnosis and HyDengue could act as a sound, efficient and smart decision-support system in early dengue prediction, which not only helps the healthcare professionals in care of patients, but also offers basis for public heath surveillance.