Early Diagnosis for Dengue Disease Prediction Using Efficient Machine Learning Techniques Based on Clinical Data

Abstract

Dengue fever is a worldwide issue, especially in Yemen. Although early detection is critical to reducing dengue disease deaths, accurate dengue diagnosis requires a long time due to the numerous clinical examinations. Thus, this issue necessitates the development of a new diagnostic schema. The objective of this work is to develop a diagnostic model for the earlier diagnosis of dengue disease using Efficient Machine Learning Techniques (EMLT). This paper proposed prediction models for dengue disease based on EMLT. Five different efficient machine learning models, including K-Nearest Neighbor (KNN), Gradient Boosting Classifier (GBC), Extra Tree Classifier (ETC), eXtreme Gradient Boosting (XGB), and Light Gradient Boosting Machine (LightGBM). All classifiers are trained and tested on the dataset using 10-Fold Cross-Validation and Holdout Cross-Validation approaches. On a test set, all models were evaluated using different metrics: accuracy, F1-sore, Recall, Precision, AUC, and operating time. Based on the findings, the ETC model achieved the highest accuracy in Hold-out and 10-fold cross-validation, with 99.12 % and 99.03 %, respectively. In the Holdout cross-validation approach, we conclude that the best classifier with high accuracy is ETC, which achieved 99.12 %. Finally, the experimental results indicate that classifier performance in holdout cross-validation outperforms 10-fold cross-validation. Accordingly, the proposed dengue prediction system demonstrates its efficacy and effectiveness in assisting doctors in accurately predicting dengue disease.