Sahay

Predictive Disaster Intelligence & Rapid Emergency Response

Advanced hydrological forecasting, rainfall analysis, blood bank locator, emergency guidelines, and incident-reporting simulation for public safety and preparedness.

Note

Sahay was originally developed in 2025 as part of the academic curriculum for the EPICS course at VIT Bhopal University. It has now been completely overhauled into a modern, dynamic web application, upgrading its model speed, repository structure, and visual aesthetics to state-of-the-art glassmorphic standards.

---

📑 Table of Contents

Click to Expand/Collapse Table of Contents

• 🌍 Overview
• ✨ Core Features
• 📸 Screenshots
• 🛠️ Technology Stack
• 📂 Project Architecture
• ⚙️ How the System Works
• 🚀 Getting Started
• ⚠️ Limitations & Assumptions
• 📚 Data Sources
• 📄 License
• 🎓 Academic Origins & Credits

---

🌍 Overview

Sahay is a comprehensive Flask-based disaster resilience platform that fuses machine-learning hydrological intelligence with practical, on-the-ground emergency support utilities. Designed with a stunning dark glassmorphic UI, it provides a unified hub to:

• Forecast river flood risks
• Analyze monthly subdivision rainfall patterns
• Locate life-saving blood banks
• Access direct-dial emergency helplines
• Study interactive disaster Standard Operating Procedures (SOPs)
• Simulate mock incident-reporting for training and demonstrations

(The active v1.0 application is housed inside app/, while the preserved v0 legacy prototype is kept under v0Archive/ to document the project's evolution.)

---

✨ Core Features

• Flood Prediction (ARIMA + LDA): Forecasts water-level metrics for major rivers (Cauvery, Godavari, Krishna, Mahanadi, Son) and classifies real-time flood risk using Machine Learning.
• Rainfall Analysis (1D-CNN + RF): Predicts April-December monthly rainfall for Indian meteorological subdivisions using either a cached Deep Learning CNN or a Random Forest Regressor.
• Blood Assistance Portal: A lightning-fast, AJAX-powered locator to find local blood bank records by state and city, complete with copy actions, direct dialers, and donor eligibility guidelines.
• Emergency Contacts: A quick-dial public safety directory for Police, Ambulance, Fire, NDRF, NDMA, and Women Helplines.
• Interactive Disaster SOPs: Actionable, tabbed do's and don'ts checklists for Floods, Earthquakes, Cyclones, Landslides, Tsunamis, Wildfires, and Thunderstorms.
• Incident Reporting Simulator: A visually engaging prototype workflow simulating emergency dispatch routing and coordinate generation.

---

📸 Screenshots

Platform Preview

Core Features

Flood Forecasting	Rainfall Analysis
Blood Bank Locator	Disaster SOPs
Incident Reporting	About

Click to view all detailed screenshots (Module-wise)

1. Home

• 
• 
• 

2. Flood Prediction

• 
• 
• 
• 

3. Rainfall Analysis

• 
• 
• 
• 
• 
• 

4. Blood Assistance

• 
• 
• 
• 

5. SOPs & Helplines

• 
• 
• 
• 

6. Incident Reporting

• 
• 
• 
• 
• 

7. About & System

• 
• 
• 
• 
• 

---

🛠️ Technology Stack

Layer	Technologies Used
Frontend UI	HTML5, Jinja2, Vanilla CSS3 (Glassmorphism), Vanilla JavaScript
Backend Framework	Python 3.10+, Flask, Werkzeug
Data Processing	Pandas, NumPy, openpyxl
Flood Models	statsmodels (ARIMA), scikit-learn (LDA), imblearn (SMOTE)
Rainfall Models	TensorFlow/Keras (1D-CNN), scikit-learn (Random Forest)
Visualizations	Matplotlib (Agg backend) for dynamic dark-mode server charting

---

📂 Project Architecture

📦 DSN3099
 ┣ 📂 screenshots/         # High-resolution UI mockups and dashboards
 ┣ 📂 v0Archive/           # Legacy v0 source code and conceptual media
 ┣ 📂 app/                 # Modern Optimized Application (v2.0)
 ┃ ┣ 📂 core/              # Isolated Machine Learning & pipeline logic
 ┃ ┃ ┣ 📜 hydrology.py     # Flood ARIMA forecasting & LDA classification
 ┃ ┃ ┗ 📜 rainfall.py      # CNN / RF rainfall prediction pipelines
 ┃ ┣ 📂 data/              # Datasets: River workbooks, IMD rainfall, Blood banks
 ┃ ┣ 📂 static/            # Frontend assets, stylesheets, and visual graphics
 ┃ ┃ ┣ 📂 css/             # CSS design system (sahay.css)
 ┃ ┃ ┣ 📂 images/          # Brand graphics and standalone SVGs
 ┃ ┃ ┗ 📂 charts/          # Dynamic chart target outputs (flood.png, rainfall.png)
 ┃ ┣ 📂 templates/         # Modular camelCase Jinja2 HTML views (base.html, index.html, etc.)
 ┃ ┣ 📂 trained/           # Serialized model artifacts (.h5, .pkl)
 ┃ ┣ 📜 __init__.py        # Flask app factory & blueprint registration
 ┃ ┗ 📜 routes.py          # Core routing, form handling, AJAX endpoints
 ┣ 📜 .env.example         # Environment configuration template
 ┣ 📜 requirements.txt     # Python package dependencies
 ┣ 📜 run.py               # Local WSGI development server entry point
 ┗ 📜 README.md            # Project documentation & specifications

---

⚙️ How the System Works

1. Flood Prediction Pipeline

User inputs River & Date ➔ hydrology.py reads historical river workbook ➔ Generates ARIMA future trend ➔ Extracts structural features ➔ Passes vector to LinearDiscriminantAnalysis ➔ Classifies Normal vs. High Risk ➔ Renders Matplotlib chart & metrics table.

2. Rainfall Analysis Pipeline

User inputs Subdivision, Year & Model ➔ rainfall.py parses IMD dataset ➔ If CNN, loads cached Keras .h5 model (or trains on-the-fly) ➔ If RF, trains RandomForestRegressor ➔ Predicts April–December mm ➔ Calculates MAE & Variance ➔ Renders comparative chart.

3. Blood Bank Search

routes.py loads blood_banks_india.csv globally on server start ➔ Normalizes strings (Title Case) ➔ Exposes /getCities and /searchBloodBanks JSON endpoints ➔ Client-side JS fetches and renders results instantly without page reloads.

---

🚀 Getting Started

1. Clone the Repository

git clone https://github.com/Varaxion/DSN3099.git
cd DSN3099

2. Create and Activate a Virtual Environment

Windows:

python -m venv epicsenv
.\epicsenv\Scripts\Activate.ps1

macOS/Linux:

python3 -m venv epicsenv
source epicsenv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Run the Server

python run.py

The Sahay portal will be live at http://127.0.0.1:5000

---

⚠️ Limitations & Assumptions

• ARIMA Horizons: Flood forecasting relies on historical moving averages; long-term horizon accuracy decays naturally.
• LDA Covariance: Flood risk classification assumes equal covariance matrices across normal and high-risk classes.
• Rainfall Granularity: Both CNN and RF models operate on aggregated monthly subdivision data, not hyperlocal or daily weather metrics.
• Incident Reporting: The reporting module is strictly a UI prototype/simulation. It does not dispatch or alert real-world NDRF agencies.
• Medical Data: Blood bank contact numbers and addresses are sourced from open datasets and should be independently verified in a real emergency.

---

📚 Data Sources

This academic project relies on several historical and open-source datasets to power its machine learning models and locator utilities:

• Hydrological River Data (Cauvery.xlsx, Godavari.xlsx, etc.): Historical water level and runoff metrics sourced from the Central Water Commission (CWC) and India-WRIS (Water Resources Information System).
• Subdivision Rainfall Data (imd-rainfall-2021.csv): Historical subdivision-wise monthly rainfall metrics provided by the Indian Meteorological Department (IMD) (frequently hosted on Open Government Data platforms).
• Blood Bank Directory (blood-banks-india.csv): Nationwide blood bank addresses and contact details aggregated from the Ministry of Health and Family Welfare (e-RaktKosh) and the Open Government Data Platform India (data.gov.in).

---

📄 License

This repository is licensed under the MIT License. You are free to use, modify, and distribute this code for your own academic or personal projects.

Warning

Exclusions & Third-Party Licenses:
The MIT License applies only to the original source code. The following assets are strictly excluded:

1. Datasets: Data files in app/data/ (CWC, IMD, e-RaktKosh) are government property and are included for demonstration purposes only.
2. Academic Documents: Reports and videos in v0Archive/docs/ are fully copyrighted by the authors and VIT Bhopal University.
3. Hyperspace Template: The legacy template in v0Archive/hyperspaceTemplate/ is governed by a separate Creative Commons Attribution 3.0 Unported (CCA 3.0) license.

---

🎓 Academic Origins & Credits

This repository originated as an academic project for VIT Bhopal University by Team-EPICS348 as part of the EPICS course in 2025. The original conceptual prototype was a collaborative effort by the following team members:

Name	Registration No.
Chelsi Patel	22BAI10005
Aditya Nayak	22BAI10424
Rushabh Wagh	22BCE10364
Kavya	22BCE10385
Tejas Pathak	22BCE10853
Simarpreet Singh	22BCE10914
Sneha Mishra	22BCE10932
Pooja	22BCE10984
Ruturaj Bhoite	22BHI10027
Archana Prasad Nair	22BSA10238

Important

v1.0 Overhaul: While the initial data gathering, research support, and legacy model iterations were developed collaboratively by Team-EPICS348, the complete system architectural overhaul, 1D-CNN model cache integration, dynamic dark-mode Matplotlib engine, responsive glassmorphic loaders, Flask app restructuring, repository documentation, and premium v1.0 CSS design system were engineered by Kavya.

Sahay - Advancing predictive intelligence for rapid emergency response.

— Re-engineered by Kavya —