(a)

• [x] Split the dataset into 70% train / 30% test. Train models with default model parameters.
• [x] Compute accuracy on training and test sets.
• [ ] Report results by creating a table with columns training accuracy, test accuracy, and AUC.

(b)

• [ ] Use 10 repeated train/test splits.
• [ ] Report the mean and standard deviation of test accuracy for each model.
• [ ] Which model has the lowest/highest variance and why?

(c)

Plot train vs test accuracy for increasing model complexity:

• [ ] For Decision Tree: vary max_depth

• [ ] For Random Forest / Bagging: vary n_estimators

• [ ] For AdaBoost: vary n_estimators

• [ ] Interpret curves in terms of bias–variance trade-off.

(d)

• draw feature importance measures (mean decrease in impurity)
• draw feature importance measures (permutation importance)
• Plot them as bar charts sorted in descending order.
• Identify the top 3 most important features for each model.
  • [ ] decision tree
  • [ ] random forest
  • [ ] AdaBoost