Support Vector Machine Outperforms Other Machine Learning Models in Early Diagnosis of Dengue Using Routine Clinical...

Area of Science:

• Epidemiology
• Infectious Diseases
• Computational Biology

Background:

• Active dengue surveillance is crucial for detecting cases, serotypes, and disease burden.
• Machine learning (ML) models offer a promising approach for predicting dengue PCR results, enhancing pre-emptive healthcare.
• This study details the data preprocessing, model selection, and mechanisms for accurate dengue PCR outcome prediction.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting dengue PCR results.
• To identify the most effective ML algorithm for dengue diagnosis.
• To document circulating dengue serotypes in Pakistan.

Main Methods:

• Analysis of data from 300 suspected dengue patients in Pakistan (August-October 2023).
• Utilized NS1 antigen ELISA, IgM/IgG antibody tests, and serotype-specific RT-PCR for dengue virus (DENV) detection.
• Employed logistic regression, XGBoost, LightGBM, random forest, SVM, and CatBoost ML models using demographic, serological, and hematological data.

Main Results:

• 184 out of 300 patients (61.33%) tested PCR positive for DENV.
• Circulating serotypes identified were DENV-1 (4.89%), DENV-2 (92.93%), and DENV-3 (2.17%).
• The Support Vector Machine (SVM) model achieved the highest performance with 71.4% accuracy, 97.4% recall, and 71.6% precision, improving to 100% recall after tuning.

Conclusions:

• The study identified three circulating dengue serotypes in Pakistan's capital territory.
• The SVM model demonstrated superior performance in predicting dengue PCR results, offering a valuable tool for rapid clinical diagnosis.
• Implementing ML models can significantly aid in dengue fever diagnosis and surveillance efforts.