Inspiration
Recent epidemics and pandemics have shown how critical early detection of outbreaks is. Often, the first signs appear as small spikes in symptoms reported across hospitals but go unnoticed until the disease spreads widely. This inspired us to build a system that can detect unusual symptom patterns early and warn authorities before an outbreak escalates.
How We Built It
We built an AI-powered pipeline using Databricks, PySpark, and Delta Lake to process anonymized hospital data containing patient pincodes and symptoms.
The system first aggregates symptom data across locations and detects abnormal spikes using an Isolation Forest anomaly detection model. When an anomaly is identified, the symptoms are analyzed using a Random Forest classifier to predict the most probable disease associated with that cluster.
The results are displayed on a Streamlit dashboard, where high-risk areas are highlighted on a map to indicate potential outbreak hotspots.
Key Features Multilingual symptom recognition using Indic AI (e.g., regional terms like “irumal” → cough) Seasonal trend awareness to distinguish normal seasonal increases from unusual spikes AI-based anomaly detection for early outbreak signals Interactive outbreak visualization dashboard
Challenges & Learning
The biggest was, collecting data. To solve this, we created synthetic data. Also, working with the Databricks platform was a bit challenging, but we managed to learn it by the end. We learnt on how to convert our idea into a pipelined working app, though on a smaller scale.
Log in or sign up for Devpost to join the conversation.