Ai Driven Predictive Analy tics for Early Disease Outbreak Detection Using Historical Multimodal Health Data

Abstract

Predicting disease outbreaks at an early stage is a critical challenge due to the
unavailability of continuous real time data and limitations associated with a ccessing paid APIs
for social media and climate information. This study aims to develop a predictive analytics
system for disease outbreak detection by utilizing previously collected and stored datasets
instead of relying on unstable API based data sources . The proposed approach integrates
historical social media reports, outbreak descriptions, and environmental factors such as
temperature and humidity, which were initially scraped during free API access periods and stored
locally for consistent use. Data p reprocessing techniques including text cleaning, normalization,
and feature extraction using TF IDF are applied to transform unstructured text into meaningful
numerical representations. The system employs Decision Tree as an existing baseline model and
Con volutional Neural Network (CNN) as the proposed model to capture complex patterns in
combined textual and environmental data. Experimental results demonstrate that the CNN model
outperforms the Decision Tree with an accuracy showing improved precision and recall in
outbreak classification. The study concludes that storing and utilizing curated datasets overcomes
API limitations while ensuring reliable and scalable disease prediction, thereby supporting early
warning systems and improving public health resp onse.