Automobile Data Visualization

The full grammar of visualization applied to one automobile spec sheet — every chart earning its place.

Deliberate visualization craft
Explicit, auditable cleaning
Exploration-to-dashboard pipeline

Scope: Full viz repertoire
Signal: Engine + weight drive price
Pipeline: Notebook to live dashboard
Craft: Chart-to-question matching

Overview

One dataset of automobile specifications — makes, body styles, engine parameters, dimensions, fuel systems, prices — pushed through the complete visualization repertoire. The point is not the cars; it is fluency. Each section of the notebook picks a real question about the market (what drives price, how engine size trades against economy, where brands position themselves) and answers it with the chart form built for that question.

Problem

Visualization skill develops unevenly when every chart comes from a different tutorial on a different dataset. Holding the data constant and varying only the visual form isolates the craft itself: with the same columns available every time, the choice of encoding becomes the entire decision — which is exactly the muscle an analyst needs.

Goals

Answer genuine market questions — price drivers, brand positioning, engine trade-offs — visually
Exercise the full seaborn/matplotlib repertoire on a single consistent dataset
Practice the discipline of matching chart form to question type

Solution

The notebook moves through question families: categorical comparisons (price by make and body style), distributions (horsepower, mileage, curb weight), relationships (engine size against price, weight against fuel economy), and multivariate views (correlation heatmaps, faceted small multiples by brand).

The recurring finding is the price story: engine size, curb weight, and horsepower carry most of the price signal, while brand adds a premium layer on top that scatter plots make visible as vertical bands within the same specification range.

This work later fed the interactive Automobile Dashboard in the UX Portfolio — the notebook is where the views were discovered; the dashboard is where the best ones went to live.

Tooling

Tech stack

Python
pandas
seaborn
matplotlib
Jupyter

Friction

Challenges

Spec-sheet data quality

Automobile datasets carry the usual real-world noise — missing entries and mixed types that arrive as strings. The cleaning pass converts and imputes explicitly so every downstream chart stands on typed, accounted-for data.

Choosing charts against habit

The temptation is to reach for the same three plots. Working through the repertoire deliberately — violin where a box plot hides shape, small multiples where a legend would overload — is what turned this from output into practice.

Takeaways

Lessons learned

Small multiples beat clever single charts for brand-comparison questions almost every time
A correlation heatmap is a table of contents for an analysis, not a conclusion
The notebook-to-dashboard pipeline works: explore wide in Jupyter, then promote the few views that survived scrutiny

Evidence

Dataset

Automobile Dataset

UCI / Kaggle

Automobile specifications — make, body style, engine parameters, dimensions, fuel system, and price — for cars from the 1985 model year.

View dataset

Roadmap

Future improvements

Refresh with a modern (post-2020) vehicle dataset including EVs
Price-prediction model to complement the descriptive story
Fold the strongest views into the rebuilt dashboard suite

Continue

UX Engineering

UX Portfolio & Dashboard Suite

Analytics

CardioGoodFitness — Customer Segmentation

Data Science

Formula 1 Data Analysis

Rebuilt for this site from the original repository — source history