Characterization of systems of thiophene and benzene with ionic liquids

Marta L. S. Batista

doi:10.1016/j.molliq.2013.10.014

Explainable Artificial Intelligence (XAI) for predicting Dengue severity using Temporal Convolution Network and Multi-modal Adaptive Fusion

2026-08-11

0

PRE

AI

T

T. Archana *

J

J. Faritha Banu

DOI：10.1007/s00521-026-12403-6

Original

Original request for help

Save

Abstract

En 中文

Dengue is a prominent global viral disease, affecting more than 60 million people annually, especially in tropical regions. Dengue is contracted through the bite of the Aedes aegypti mosquito, with symptoms of fever, rash, headache and vomiting, while advanced cases can also develop shock and organ failure. Several machine learning approaches to the analysis of dengue exist, but most studies primarily focus on case forecasting or single-modal data, with limited capability to accurately predict disease severity and insufficient model interpretability. Therefore, there is a need for robust, integrated and multimodal framework for the early prediction of dengue severity. This study integrates Temporal Convolutional Networks (TCNs) and a Multi-Modal Adaptive Fusion (MAF) mechanism to predict three levels of dengue severity: mild, moderate and severe. The TCN effectively captures long-term temporal dependencies from climate variables, while MAF fuses climatic, clinical and demographic features. Kaggle dengue clinical records and India daily weather data are utilized to improve the differentiation of case severity. Furthermore, SHAP-based explainability is incorporated to provide transparent insights into key predictive factors influencing severity outcomes. The proposed framework achieved 98.5% accuracy, 97.8% precision, 97.2% recall and an AUC-ROC of 0.98. The high predictive performance and the explainable feature attribution emphasise the effectiveness of the proposed TCN-MAF framework. This study offers a clinically meaningful and interpretable decision-support tool for early dengue severity prediction, facilitating timely intervention and informed medical decision-making.

Keywords:

Dengue haemorrhagic fever (DHF)

Dengue shock syndrome (DSS)

Temporal convolutional networks (TCNs)

Multi-modal adaptive fusion (MAF)

SHapley additive explanations (SHAP)

Explainable AI (XAI)

Explainable Artificial Intelligence (XAI) for predicting Dengue severity using Temporal Convolution Network and Multi-modal Adaptive Fusion

Abstract

Journal

Organization

Cited Papers

Citing Papers