Title : Applying machine learning and deep learning for multiclass classification of arboviral diseases: A comparative analysis
Abstract:
I. MOTIVATION
Arboviral diseases, a significant subclass of Neglected Tropical Diseases, impose a considerable health burden globally, particularly in lower-income nations. Dengue, Chikungunya, and Zika trio account for a high proportion of morbidity and mortality in Latin America, South East Asia, and South America. The task of early diagnosis is complex due to symptom similarities and serologic test cross-reactions. Resource limitations, including testing kits, exacerbate the situation in remote locations, hindering timely diagnosis and management. Machine learning (ML) has recently been employed to aid clinicians in distinguishing these arboviral diseases. However, previous research primarily focused on binary classification using ML algorithms, with scant attention given to multi-class disease classification. To our knowledge, prior studies have yet to explore the application of deep learning (DL) algorithms for the multi-classification of arboviral diseases.
II. HYPOTHESIS
ML and DL algorithms can effectively facilitate multi-class classification of arboviral diseases using clinical and epidemiological data.
III. METHODS AND RESULTS
This research uses clinical and socio-demographic data to evaluate the efficacy of diverse ML and DL algorithms for the multi-class classification of Dengue, Chikungunya, and inconclusive cases. The Synthetic Minority Oversampling Technique (SMOTE) was implemented to address imbalanced data.
The models were also validated on an external testing dataset. The results reveal that the Random Forest algorithm offered superior performance, achieving an Area Under the Curve of Receiver Operating Characteristic (ROC AUC) of 99% on the validation set and 83% on the external test set. DL models accomplished a ROC AUC of 86% and 87% on the validation and external test sets, respectively.
IV. CONCLUSION
This study demonstrated the applicability and impressive performance of ML and DL algorithms in the multi-class classification of arboviral diseases using exclusively clinical and epidemiological data. Our models hold promise for supporting clinicians in distinguishing arboviral diseases, especially in resource-constrained environments where laboratory-based diagnoses are not feasible.
Keywords – Arboviral diseases, Clinical support system, Deep learning, Machine learning, Multi-class classification
Audience Take Away:
- Provide an automated clinical decision support system for differentiating diagnosis of dengue and Chikungunya disease
- This research could be applied in other infectious diseases
- This research provide a practical solution to the arboviral diseases diagnosis problem, especially in resource-limited areas
