This project offers a comprehensive, data-driven tool for understanding and predicting disease patterns in India. It moves beyond simple prediction by integrating real-world, location-based data and providing powerful geospatial visualizations to uncover hidden trends.
The core mission is to create an accessible web application that serves two primary functions: first, to provide an intelligent, context-aware disease prediction based on symptoms, and second, to identify and visualize endemic hotspots of various diseases across the country.
-
Intelligent Prediction: To develop a robust machine learning model that accurately predicts potential diseases from a set of user-provided symptoms.
-
Contextual Accuracy: To significantly enhance the prediction model's relevance by factoring in real-world disease prevalence data from the user's specific state.
-
Geospatial Insight: To apply unsupervised clustering techniques to a large-scale dataset to automatically identify and map geographical disease hotspots.
-
User-Friendly Interface: To wrap these complex analytical tools in a simple, interactive, and easy-to-use Streamlit web application.
The application's functionality is split into two sophisticated components: a Predictive Model and a Geospatial Cluster Analysis map.
The prediction feature is more than just a simple symptom checker. It employs a two-stage process to deliver a nuanced and realistic assessment.
-
Stage 1: Machine Learning CoreA Random Forest Classifier serves as the predictive engine. This powerful ensemble model, composed of hundreds of individual decision trees, is trained on a vast dataset linking symptoms to diagnoses. It analyzes the user's input symptoms to calculate a baseline probability for a wide range of diseases.
-
Stage 2: Location-Aware ScoringThis is the project's key innovation. The model understands that a disease's likelihood is not just about symptoms—it's also about location. The baseline probabilities from the Random Forest are intelligently adjusted using real-world active case data. The final score for each predicted disease is a weighted combination of the model's symptom-based confidence and the actual prevalence of that disease in the user's selected state. This makes the predictions far more practical and grounded in reality.
The "Interactive Map" tab reveals macro-level insights about disease distribution across India.
-
Unsupervised Clustering: The system uses the K-Means clustering algorithm on a dataset containing geo-coordinates (latitude/longitude), disease diagnoses, and environmental factors. K-Means is an unsupervised method, meaning it finds natural patterns and groups in the data without being told what to look for.
-
Visualizing the Results: The algorithm groups the data into 10 distinct clusters. Each cluster represents a geographical area with a unique disease profile. These clusters are then plotted on an interactive Folium map, with each cluster assigned a different color. The map features a detailed legend that automatically analyzes and describes each cluster's characteristics, such as "Plains Dengue & Malaria Zone (Monsoon)"—transforming raw data into actionable geographical intelligence.