Predictive Modeling of Auto MPG and Analysis of Engine Design Factors: Comparing the Impact of Weight and Displacement

Author: Seonwoo Kang

Field of Study: Mathematical Sciences

Program Affiliation: Summer Scholars and Artists Program (SSAP)

Faculty Mentors: Dang Huy

Easel: Digital Presentation

Timeslot: Morning

Abstract: This research explores technical determinants of vehicle fuel efficiency by developing a Multiple Linear Regression (MLR) model using the 398-instance Auto MPG dataset. It evaluates the relationship between miles per gallon (MPG) and specifications like displacement, weight, and technological progress. To address high correlation between displacement and weight (r=0.93), the model uses standardized coefficients to isolate variables. The model achieves high predictive accuracy (R^2=0.8244, RMSE=3.0725), explaining over 82% of fuel efficiency variance. Analysis identifies vehicle weight as the most significant negative factor, with an impact nearly twice that of technological gains. Furthermore, a positive model year coefficient confirms engineering progress improves economy regardless of physical dimensions. Findings indicate that prioritizing weight reduction and continuous technological evolution is mathematically more effective for energy economy than displacement downsizing alone.