ranazsaad/QWeather_Forcasting

Quantum Dry-Run Hackathon project exploring quantum machine learning for temperature prediction in Cairo, Egypt # Quantum Weather Forecasting 🌩️☔ This project was submitted for the **Quantum AI Hackathon** by **Team 5**. ## Team Members - Rana Saad - Sherifa Helmy - Yousef Ahmed - Jana Mohamed --- ## Project Overview This project explores the potential of quantum machine learning (QML) for temperature forecasting by utilizing a dataset containing daily weather observations for Cairo, Egypt We implemented a **Quantum Neural Network** using Qiskit Machine Learning and benchmarked it against classical linear regression, achieving remarkable results with only 2 key features. ## Step-by-Step Breakdown ### 1. Classical Data Preprocessing - Removed irrelevant/duplicate columns - Identified identical columns and dropped them - standard scaling is applied to the input features --- ### 2. Classical Baseline Model We implemented a Multiple Linear Regression model - Used a correlation heatmap to analyze feature relationships and selected the top 8 features that were strongly correlated - Split data into training and test sets (80/20). - Trained a Multiple Linear Regression model. ### Classical Model Results Serves as performance benchmark for our quantum approach - **MSE**: 0.0738 - **R²**: 0.9981 --- ### 3. Quantum Data Preprocessing - Removed irrelevant/duplicate columns with quantum efficiency in mind - Selected top 2 most impactful features for quantum processing: - `apparent_temperature_mean (°C)` - `et0_fao_evapotranspiration (mm)` - Applied specialized MinMax scaling (-1 to 1) optimal for quantum feature maps --- ### 4. Quantum Neural Network Implementation ⚛️ **Breakthrough Quantum Architecture:** - Used `Qiskit` and `qiskit-machine-learning` with PyTorch integration - Quantum circuit components: - `ZZFeatureMap` for optimal quantum feature encoding - `EfficientSU2` (with circular entanglement) for variational optimization - Constructed **Hybrid Quantum-Classical Model**: - Quantum layer via `EstimatorQNN` - Classical post-processing with PyTor …