Enhacing Air Quality Index Prediction: A Comparatice Analysis of Transformer Models, Graph Neural Networks, and Traditional Approaches

No Thumbnail Available
Sharma, Aditya
Issue Date
MSc in Business Analytics
Dublin Business School
This study presents a comprehensive comparative analysis of air quality index (AQI) prediction models, focusing on Transformer Models, Graph Neural Networks (GNN), and traditional approaches such as Linear Regression, Naive Bayes, and Long Short-Term Memory (LSTM). Evaluation metrics include Mean Squared Error (MSE) and Root Mean Squared Error (RMSE). Among the models, Graph Neural Networks exhibit superior performance with an MSE of 1601.86 and an RMSE of 40.02. The Hybrid Model follows closely with an MSE of 2057.53 and an RMSE of 45.36. Traditional methods like Linear Regression and Naive Bayes demonstrate moderate accuracy, while the LSTM model exhibits higher errors. Notably, the Transformer Model records the highest errors, suggesting challenges in accurately predicting AQI using this approach. These findings offer valuable insights for selecting optimal models to enhance air quality predictions in environmental research and monitoring.