Air Quality Dataset for Green AI-Based Machine Learning Evaluation
Description
This dataset is the UCI Air Quality dataset used in a Green AI research study for evaluating machine learning models with respect to predictive performance and computational sustainability. The dataset contains hourly averaged air quality measurements collected using a multisensor device and reference analyzer. It is used in experiments comparing regression models based on R², RMSE, MAE, training time, estimated energy consumption, estimated carbon emissions, and a Green AI Efficiency Score. The original dataset was created by Saverio Vito and is available through the UCI Machine Learning Repository. The original dataset is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. Original source: Vito, S. (2008). Air Quality. UCI Machine Learning Repository. DOI: 10.24432/C59K5F.
Files
Steps to reproduce
Load the air quality dataset into the Green AI Performance Analyzer. Perform the required data preprocessing and prepare the features and target variable for regression. Split the data using the chronological train-test evaluation procedure described in the accompanying manuscript. Train the selected regression models on each evaluation split. Calculate R², RMSE, MAE, training time, estimated energy consumption, estimated carbon emissions, and the Green AI Efficiency Score for each model. Repeat the evaluation across the five chronological splits and aggregate the results to compare predictive performance and computational sustainability.