9.3 Titanic Data Example

Create Interpretable Data Space

• Combine similar concepts
  • group class and fare into “wealth”
    
  • combine age and gender into “demographics”
    
• Simpler approach for this example - discretize every variable into two levels
  • age bucketed into “<= 15.36” and > 15.36
  • class into “1st/2nd/deck crew” and “other”

LIME Applied to Johnny D

Figure 9.5: Johnny D LIME, LASSO, K=3