Air pollution is a significant environmental concern that affects the health and well-being of millions of people around the world. Understanding the underlying factors and relationships between various pollutants is crucial for devising effective interventions and improving air quality. This project aims to explore the causal relationships among different air quality indicators using an Invariant Variational Autoencoder (iVAE).
Traditional correlation-based methods can struggle to identify true causal links due to the complex and intertwined nature of environmental data. By leveraging iVAE, this project adopts a modern approach to causal inference that aims to reveal invariant relationships between pollutants under different conditions, which are often masked by noise or hidden variables.
The primary objectives of this project are to:
- Extract causal features from air quality time series data.
- Identify causal relationships between key pollutants and environmental factors, such as temperature.
- Demonstrate the use of iVAE as a tool for performing causal analysis in environmental studies.
This project uses publicly available air quality data and offers insights into how advanced machine learning techniques can help understand the interplay between various pollutants and environmental conditions. Through this work, we hope to contribute to ongoing research in environmental health, machine learning, and causal inference.