CarLab is an interactive R Shiny web app for comparing cars using:

1. a preference-based FitScore ranking, and
2. a multivariate PCA + k-means clustering view to understand performance trade-offs.

Live app: https://markchweya.shinyapps.io/carlab/

• Set your priorities (dropdown levels):
  • Fuel economy (mpg)
  • Engine power (hp)
  • Lightweight (lower wt)
  • Fast acceleration (lower qsec)
• Apply practical filters:
  • Cylinders
  • Transmission (Automatic/Manual)
  • Minimum mpg
• Instantly get:
  • A Best match car card
  • A Top 10 ranked list you can tap to open a zoom modal

How FitScore is computed

• Each feature is standardized (z-score).
• Weight and acceleration are flipped so “better” is higher (-wt, -qsec).
• Dropdown priorities map to weights: Low=1, Medium=2, High=3, Very high=4.
• Final score = weighted sum of standardized features.

• Choose which variables to include in PCA (mpg, hp, wt, qsec, drat)
• View a PC1 vs PC2 performance map
• Run k-means clustering (k = 2–6) on PC1–PC2
• Optional car labels
• View supporting tables:
  • Explained variance (eigenvalues, proportions, cumulative)
  • Feature influence (PCA loadings)
  • Cluster profiles (avg mpg, hp, wt, qsec, cars per cluster)
• Click a point (car icon) to zoom into a car modal with details

• Uses the built-in R dataset: mtcars
• Adds helper fields:
  • car (name from rownames)
  • transmission (Automatic/Manual from am)
  • cylinders (factor from cyl)

• R
• Shiny
• ggplot2 (visuals)
• dplyr (data wrangling)
• DT (raw data table modal)
• bslib (theme + styling support)

• “Hologram / Vision Pro” glassmorphism cards
• Parallax motion layer (pure JS)
• Rotating “wheel text” feature callouts on the Home orb
• Tap-to-zoom car cards + plot click-to-zoom modal

install.packages(c("shiny", "ggplot2", "dplyr", "DT", "bslib"))