milan-air-quality-forecast-lstm

Milan Air Quality Forecast using LSTM

Forecasting daily PM10 air pollution levels in Milan using Long Short-Term Memory (LSTM) deep learning models.

Project Status: ✅ Completed

Project Overview

This project develops a Long Short-Term Memory (LSTM) deep learning model to forecast daily PM10 concentrations in Milan, Italy. The objective is to predict future air pollution levels using historical PM10 observations and evaluate the effectiveness of LSTM for time-series forecasting.

Objectives

• Collect PM10 air quality data
• Prepare and clean the dataset
• Build an LSTM forecasting model
• Evaluate forecasting performance
• Predict future PM10 concentrations
• Compare forecasting results

---

Dataset

The project uses historical PM10 air quality measurements for Milan.

Processed datasets include:

• Daily PM10 observations
• Cleaned datasets
• Prepared LSTM sequences

---

Project Workflow

1. Download air quality data
2. Extract Milan PM10 observations
3. Clean missing values
4. Prepare daily time series
5. Create LSTM training sequences
6. Train the LSTM model
7. Evaluate model performance
8. Generate future forecasts
9. Compare forecasting results
10. Visualize predictions

---

Technologies Used

• Python
• Pandas
• NumPy
• TensorFlow / Keras
• Matplotlib
• Scikit-learn
• Jupyter Notebook

---

Installation

git clone https://github.com/MuhammadFaisalHaroon/milan-air-quality-forecast.git
cd milan-air-quality-forecast

---

Repository Structure

├── data/
├── notebooks/
├── outputs/
├── models/
├── src/
├── README.md
└── LICENSE

---

Results

The project successfully:

• Prepared historical PM10 data
• Trained an LSTM forecasting model
• Generated future PM10 predictions
• Evaluated forecasting accuracy
• Produced visualizations and comparison results

The project demonstrates the effectiveness of LSTM networks for PM10 time-series forecasting and provides a complete end-to-end machine learning workflow.

---

Future Improvements

• Experiment with Transformer models
• Compare with GRU and XGBoost
• Include meteorological variables
• Deploy as a Streamlit web application

---

Author

Muhammad Faisal Haroon

GitHub Profile: MuhammadFaisalHaroon

GitHub Repository: milan-air-quality-forecast