Multi-Agent AI Framework for the Early Diagnosis of Tuberculosis

Abstract

Stunting is a critical public health issue, particularly in developing countries, with long-term impacts on physical and cognitive development due to chronic malnutrition. In the digital era, public sentiment, and opinion on stunting, as reflected on social media platforms, can provide valuable insights for policymakers and healthcare providers. However, there has been limited research combining multiple machine-learning algorithms for sentiment analysis on this topic. This study aims to conduct a comparative analysis of user sentiment on X regarding public opinion on stunting using three widely utilized text classification algorithms: K-Nearest Neighbour (KNN), Naïve Bayes, and Decision Tree. The research involves collecting and annotating sentiment data (positive, negative, neutral) from social media platforms like Twitter and X, followed by model training and evaluation using metrics such as accuracy, precision, recall, and F1-score. Results from the analysis of 1227 data points (2023-2024) with an 80:20 train-test split revealed that Naïve Bayes achieved the highest accuracy at 87.18%, followed by Decision Tree at 79.83%, and KNN at 57.56%. Meanwhile, a comparison using a 70:30 train-test split yielded similar results, with Naïve Bayes achieving 86.41%, Decision Tree 78.41%, and KNN 54.48%. The study demonstrates the effectiveness of the Naïve Bayes algorithm for sentiment analysis in public health contexts, offering a novel approach to understanding public opinion on stunting. By leveraging social media data, this research provides real-time insights that can guide policy and public health interventions. Furthermore, the comparative analysis of algorithms provides valuable knowledge on the most effective models for sentiment analysis in complex, sentiment-driven public health issues. The study emphasizes the role of machine learning in addressing malnutrition and stunting through data-driven policy decisions.