Automated detection and prediction of dengue fever: A systematic review from 2013 to 2025

Abstract

Timely diagnosis and forecasting of dengue fever cases help to control its spread and lessen its negative effects on public health. Modern advancements in artificial intelligence (AI) made it possible to use automatic approaches to diagnose, predict, classify, monitor and assess the risks of dengue fever. Many Machine learning (ML) and deep learning (DL) algorithms are used in the current research into dengue to analyze and work with data on patients, disease outbreaks, the environment, spatial information, and images. However, many issues remain to be solved concerning data quality, class imbalance, validation approaches, interpretability, replicability, and implementation of ML and DL solutions. This review summarizes the usage of ML and DL algorithms in diagnosing, predicting, classifying, and assessing the risks of dengue fever, relying on the relevant articles available in the literature from 2013 to 2025. It is guided by the Preferred Reporting Items for Systematic Reviews and MetaAnalyses (PRISMA) criteria and identifies articles in the IEEE Xplore, Scopus, Web of Science, and PubMed databases. Task-oriented applications of ML and DL in studies on dengue fever are presented along with data modalities, validation techniques, methods' limitations, and reported performance results. Research findings show that random forest and extreme gradient boosting are often used to predict and classify dengue outbreaks, whereas convolutional neural networks (CNNs) and long short-term memory (LSTM) algorithms demonstrate great potential in detecting images and temporal forecasting of dengue fever. Nevertheless, metrics have to be considered carefully since there is significant heterogeneity between studies in data sets, targets, class imbalance, preprocessing, and validation.