Abstract:
Fever and has been reporting outbreaks of the disease, it still continues to remain an important public health concern of the new millennium with significant morbidity strains on the healthcare system. It is important for early detection and error-free prevention of the disease. The included subjects comprised a group of patients admitted to multiple various healthcare centers linked with medical records (i.e. clinical patient data (gender, age), hematological data and other conventional clinical indicators). The dataset had been pre-processed to treat missing values and outliers in data provided, scaled for model learning. The models that are used in this study is Random Forest, Logistic Regression, Decision Tree, Support Vector Machines (SVM), Neighbors (KNN) K-Nearest, GaussianNB, XGBoost Stacking Ensemble. The study results indicated that Random Forest achieved the highest accuracy (0.89), followed closely by XGBoost (0.87) and the Stacking Ensemble (0.88). Decision Tree and Support Vector Machines (SVM) demonstrated moderate performance with accuracies of 0.82 and 0.83, respectively. Logistic Regression performed fairly well at 0.79, while comparatively lower accuracies were observed for K-Nearest Neighbors (KNN) at 0.76 and Gaussian Naïve Bayes (GaussianNB) at 0.74. The models were evaluated based on precision, recall, F1 score and confusion matrices where Random Forest and Logistic Regression performed the best in terms of minimizing false positives and false negatives. Finally, we have showed being useful machine learning models, in this specific Random Forest and Logistic Regression in the prediction of outbreaks of dengue. However, the results of this analysis suggest that the use of such models maybe beneficial for more efficient monitoring of surveillance, enhancing resource allocation and thus more timely interventions in controlling Dengue Fever in Bangladesh. Higher quality of real-time model parameter estimations with more types of input information may improve the reliability and support to decision makers.
