Climate Change Impact Predictor

A comprehensive portfolio project building a global climate change impact prediction system with advanced machine learning and real-time data pipeline integration.

Project Overview

This project demonstrates end-to-end data science and machine learning engineering skills by:

• Advanced ML Architecture: LSTM time series forecasting + Multi-output deep learning
• Global Scale: Trained on 150+ world capitals with real-time predictions for any location
• Production Ready: Professional code architecture with comprehensive error handling
• Multi-Model Ensemble: 3 specialized ML models working together

Quick Start

Installation

# Clone the repository
git clone https://github.com/YOUR_USERNAME/climate-change-impact-predictor.git
cd climate-change-impact-predictor

# Set up virtual environment
python -m venv climate_env
source climate_env/bin/activate  # On Windows: climate_env\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Running the Web Interface

# Activate the virtual environment
source climate_env/bin/activate

# Start the ML API server (required for predictions)
PYTHONPATH=$(pwd) python app/api_server.py &

# Start the Streamlit web interface
PYTHONPATH=$(pwd) streamlit run app/location_picker.py

The web interface will be available at: http://localhost:8501 The ML API will be running at: http://localhost:8000

Running CLI Predictions

# Single city prediction
PYTHONPATH=$(pwd) python -m tools.day10_advanced_demo Lahore

# Interactive demo
PYTHONPATH=$(pwd) python -m tools.day10_advanced_demo

# Multiple cities
PYTHONPATH=$(pwd) python -m tools.day10_advanced_demo "New York"
PYTHONPATH=$(pwd) python -m tools.day10_advanced_demo "Tokyo"

Python API Usage

from src.models.enhanced_climate_predictor import EnhancedGlobalClimatePredictor

# Initialize enhanced predictor
predictor = EnhancedGlobalClimatePredictor()

# Basic prediction
result = predictor.predict_climate("London")

# LSTM long-term forecasting
forecast = predictor.predict_long_term("Tokyo", days=30)

# Multi-output comprehensive prediction
comprehensive = predictor.predict_comprehensive("Berlin")

# Ultimate prediction (all models combined)
advanced = predictor.predict_climate_advanced("Paris")

Data Sources & Coverage

API Source	Data Type	Coverage	Purpose
Open-Meteo	Air Quality & Weather	Global hourly	Real-time pollution monitoring & forecasts
NASA POWER	Meteorological	Global daily	Historical weather foundation
World Bank CCKP	Climate Projections	Country-level	IPCC CMIP6 future scenarios

Dataset Scale: 150+ world capitals, 60+ climate features, multi-source validation

Machine Learning Models

1. Base Climate Model

• Architecture: Feed-forward neural network
• Training: 144 world capitals
• Capability: Basic climate prediction for any city

2. LSTM Time Series Forecaster

• Architecture: Bidirectional LSTM with attention mechanisms
• Training: Enhanced dataset with time series sequences
• Capability: 7-30 day weather forecasting

3. Multi-Output Deep Learning

• Architecture: Multi-head dense networks with shared feature extraction
• Training: Comprehensive multi-target learning
• Capability: Simultaneous prediction of 5+ climate variables
• Outputs: Temperature, precipitation, UV index, air quality, wind

System Architecture

Climate Change Impact Predictor
├── Global Location Service (any coordinates)
├── Multi-Source Data Pipeline (3 professional APIs)
├── Advanced ML Ensemble (3 specialized models)
├── Enhanced Prediction Engine (production-ready)
├── REST API Server (FastAPI)
└── Interactive Web Dashboard (Streamlit)

Project Structure

├── app/                       # Web application
│   ├── api_server.py          # FastAPI ML prediction server
│   └── location_picker.py     # Streamlit web interface
├── data/                      # Multi-source climate datasets
│   ├── raw/                   # Raw API responses
│   ├── processed/             # Processed and integrated data
│   ├── capitals/              # World capitals training data
│   └── day10/                 # Advanced ML training datasets
├── models/                    # Trained ML models and preprocessors
│   ├── global_climate_model.keras
│   ├── advanced_lstm_forecaster.keras
│   └── production_multi_output_predictor.keras
├── notebooks/                 # Training and analysis notebooks
├── src/                       # Core system implementation
│   ├── api/                   # Multi-API integration layer
│   ├── core/                  # Data management and processing
│   ├── features/              # Universal feature engineering
│   ├── models/                # ML model implementations
│   └── validation/            # Quality assurance and testing
└── tools/                     # Professional tooling and demos

Development Status

Phase 1: Foundation & Data Pipeline (Complete)

• Multi-API integration (Open-Meteo, NASA POWER, World Bank CCKP)
• Advanced data processing pipeline with 60+ climate features
• Universal location service supporting any global coordinates
• Adaptive data collection with smart source selection

Phase 2: Advanced Machine Learning (Complete)

• Global neural network trained on 144 world capitals
• LSTM Time Series Forecasting for 7-30 day predictions
• Multi-Output Deep Learning for simultaneous predictions

Phase 3: Web Application (Complete)

• REST API server for model serving
• Interactive Streamlit dashboard
• Real-time climate predictions

API Endpoints

Endpoint	Method	Description
/health	GET	Check API health status
/predict/basic	POST	Basic climate prediction
/predict/forecast	POST	LSTM weather forecast
/location/search	GET	Search for locations

Requirements

• Python 3.10+
• TensorFlow 2.x
• Streamlit
• FastAPI
• See requirements.txt for full dependencies

License

MIT License - See LICENSE file for details.

---

This project showcases advanced machine learning engineering skills with real-world climate data and production-ready architecture.