CarPrice

Kaggle > Notebook

Decision Tree Regression with Used Car Prices in R (2022.12.07)

• Sources

  • Kaggle Dataset : used car price dataset (competition format)

  • Kaggle Notebook : DT(rpart) Practice with Used Car Prices in R

• Pre-processing

  • The given data has been already divided into Train/Test data and there's no missing data(:heart_eyes:)

  • Remove 3 variables : carID brand model

• Model fitting : Use rpart library in R and try Decision Tree Regression

  • At the 1st trial, there's no point for xerror to rise up again, but I feel I should do something …… (:scream:)

  • But at the 2nd trial, it shows rather worse performance. (:sob:)

• Compare 3 models with different CP(Complexity Parameter) values

  models	cp	nsplit	min_xerror	correlation	rmse
  1	0.010 (default)	10	0.2629	0.8432	8951
  2	0.025	5	0.3538	0.7950	10093
  3	0.001	45	0.1484	0.9021	7230

  

• Takeaway

  • the result from rpart() is not a regression formula, but it just outputs some "countable" kinds of values. If I knew it, I wouldn't try cutting cp hastly.