Air Quality Analysis and Prediction

This project focuses on analyzing and predicting air quality using machine learning techniques. Leveraging a detailed dataset containing air pollutant concentrations, temperature, and humidity readings, the project builds and evaluates predictive models, including LSTM and RNN architectures, to forecast air quality metrics.

Dataset

The project utilizes the Air Quality UCI Dataset, which contains 9,357 rows and 15 columns. Key attributes include:

• Date and Time: Timestamps of measurements.
• Gas Concentrations: CO, NMHC, NOx, NO2, and others.
• Sensor Outputs: Measurements from various air quality sensors.
• Environmental Parameters: Temperature (T), Relative Humidity (RH), and Absolute Humidity (AH).

The dataset provides a comprehensive view of air quality indicators over time.

Project Features

1. Data Preprocessing:

   • Handling missing values and outliers.
   • Normalizing data for model readiness.
   • Feature engineering to enhance model performance.

2. Machine Learning Models:

   • Implementation of Long Short-Term Memory (LSTM) and Recurrent Neural Networks (RNN) for time-series forecasting.
   • Use of Keras to design and train deep learning models.

3. Evaluation Metrics:

   • RMSE, MAE, and R² scores to assess model performance.
   • Comparison of results between LSTM and RNN approaches.

Repository Contents

• Dataset: Includes the processed Air Quality UCI dataset.
• Notebook: Step-by-step analysis and modeling in amritdev_project3.ipynb.
• Models: Pre-trained LSTM and RNN models provided in .h5 and .keras formats.

How to Run

1. Clone the repository:

   git clone https://github.com/yourusername/air-quality-analysis.git

2. Install the required Python libraries:

   pip install -r requirements.txt

3. Open the Jupyter notebook to explore data preprocessing and modeling:

   jupyter notebook amritdev_project3.ipynb

4. Use the pre-trained models for predictions or train new models using the provided notebook.

Applications

This project can be utilized for:

• Real-time air quality monitoring.
• Policy-making to improve environmental health.
• Research on the impact of pollutants on urban areas.

Future Work

• Integration with real-time air quality sensors.
• Deployment of the models in a web application.
• Exploration of other deep learning architectures for improved accuracy.

License

This project is licensed under the MIT License. See the LICENSE file for details.