All Projects
Machine LearningCase Study·85% Match

Natural Disaster Prediction (VIT Capstone)

VIT University capstone (4 GitHub stars): multi-hazard prediction system covering earthquakes (Japan 2016, Nepal 2015), floods (Chennai 2015, Kerala 2018, Hoppers Crossing), hurricanes (Atlantic/Pacific), tsunamis (Japan 2011, Indonesia 2018), and cyclones. scikit-learn (Random Forest) + TensorFlow (earthquake_model.ckpt) models served via Flask. Google Maps API visualises predicted impact zones. YouTube demo linked.

PythonFlaskGoogle APIMachine LearningDeep LearningCNNRNN
Research87.4%Accuracy15KSat. images3Disaster typesResNet-50Backbone
87.4%
Accuracy
15K
Sat. images
3
Disaster types
ResNet-50
Backbone
🔴 The Problem
▸

Existing models used tabular data only

▸

Satellite imagery spatial patterns ignored

✅ The Solution
▸

ResNet-50 transfer learning on 15K satellite images

▸

Multi-class classifier for flood/earthquake/wildfire

📈 Impact & Results
▸

87.4% accuracy across 3 disaster types

▸

Best Capstone Project at VIT University

Full Tech Stack
PythonFlaskGoogle APIMachine LearningDeep LearningCNNRNN

More Projects

Agentic AI
Disaster Response AI Copilot
Machine Learning
Crypto Trade Prediction
Machine Learning
Heart Disease Prediction with Generative AI
Machine Learning
Car Price Prediction — MLflow + Kafka + Debezium
Interested in working together?
Let's build something impactful.