UChicago Pathways in Data Science · July 2026
Automobile Efficiency Across Eras began as my team’s final project in a three-week summer program. After I returned home, I kept the original question and built outward: more decades, more careful tests, more transparent data notes, and a public site where readers can examine the work directly.
01 · The handoff
The most important part of the project was not the final presentation. It was realizing how many good questions remained after it.
At UChicago
I coordinated data cleaning, model comparison, interpretation, and presentation for a four-member team's study of nearly 400 historical automobiles. We used k-nearest neighbors and multiple linear regression to predict miles per gallon.
Back at home
I kept the historical analysis, added 1999–2008 vehicle data and EPA trends through 2024, separated sources with different testing rules, and built new studies of long-run trends, chronological prediction, cross-era model drift, and model-adjusted residuals.
02 · What grew
I did not force every source into one oversized table. Each dataset has a specific role, and the project keeps those boundaries visible.
The original UCI file supports historical model comparison. EPA annual data supplies long-run, production-weighted context. Later vehicle files make chronological and cross-era tests possible.
The expanded analysis compares ordinary train/test results with later-year holdouts, held-out model families, and transfer tests between vehicle periods. Accuracy declined on newer cars, showing how SUV and pickup growth limited overall efficiency gains.
Six notebooks, a repeatable data pipeline, a written report, source documentation, and a route-verified static site now carry the work beyond the original classroom presentation.
One result in context: EPA real-world fuel economy rose from 13.1 MPG in 1975 to 27.2 MPG in 2024. The later-year tests also showed why that overall improvement does not mean a model trained on older vehicles will transfer cleanly to newer ones.
03 · Follow the work
Each part of the analysis has a separate page so a reader can go directly to the question that matters to them.
Compare source-specific models and try the project’s predictors.
Trace fuel economy, vehicle mix, horsepower, weight, and technology over time.
See what changes when a model is carried from one vehicle period into another.
Review source coverage, testing-rule differences, and the limits of comparison.
What I carried home