Estimating Corporate Greenhouse Gas Emissions

This paper explores machine learning models for estimating scope 1, 2, and 3 greenhouse gas emissions of companies using commonly available data. We scrutinize each preprocessing step and propose advanced solutions for outlier filtering, out-of-sample evaluation, missing value imputation, and predictor selection. We suggest two frameworks: one without restriction and another one with fewer inputs, better suited for small and mid-caps and private companies.