CodeCheck : how do our food choices affect climate change?

Abstract

Different approaches were proposed to predict the carbon footprint of products from the different datasets provided by CodeCheck. Multivariate linear regression and random forest regression models perform well in predicting carbon footprint, especially when - in addition to the nutrition information - the product categories, learned through Latent Dirichlet Allocation (LDA), were used as extra features in the models. The prediction accuracy of the models that were considered varied across datasets. A potential way to display the footprint estimates in the app was proposed.