PREDICTING BUILDING ENERGY EFFICIENCY IN CHICAGO: A PREDICTIVE ANALYTICS APPROACH

Abstract

Urban centers worldwide, including Chicago, are striving to reduce energy consumption as a critical part of their sustainability efforts. In Chicago, buildings account for approximately 70% of the city's total carbon emissions, underscoring the urgent need to enhance building energy efficiency. This study develops predictive models to identify factors influencing energy performance in large buildings (≥50,000 square feet) using Chicago's Energy Benchmarking dataset from 2014-2023. Through comprehensive analysis of 16,495 buildings using multiple modeling approaches including Ordinary Least Squares (OLS), Weighted Least Squares (WLS), Partial Least Squares Regression (PLSR), and Random Forest, we identified floor area, greenhouse gas emissions, source and site energy use intensity, natural gas use, and property type as significant predictors of Energy Star Scores. The Random Forest model achieved the best predictive performance (RMSE = 11.09), while the linear-log OLS model (RMSE = 19.01, R² = 0.52) provided superior interpretability. Results indicate that larger buildings benefit from economies of scale in energy efficiency, and targeted interventions to reduce greenhouse gas emissions and site energy use intensity could significantly improve building energy performance. This research provides actionable insights for city leaders, sustainability officers, and property owners to make data-informed decisions, prioritize retrofits, and support Chicago's climate goals.