Steepest and Flattest — medium Machine Learning in Code problem | Incognition

Steepest and Flattest

10 pts · 10 coins

Machine Learning in CodeMedium

Steepest and Flattest

A forest predicts diabetes progression. For a handful of columns, which one moves its predictions the most, and which barely moves them at all?

Measure each curve's size as its range: the highest average prediction minus the lowest.

Task: write steepest_and_flattest(columns).

Load load_diabetes(return_X_y=True, as_frame=True).
Split with train_test_split(X, y, test_size=0.25, random_state=0).
Fit RandomForestRegressor(n_estimators=50, random_state=0) on the training rows.
For each name in columns, run partial_dependence(model, X_train, features=[name], grid_resolution=10) and take its ["average"][0].
Return [steepest, flattest]: the name with the biggest range and the name with the smallest.

Examples

columns = ["bmi", "age", "bp"]

returns ['bmi', 'age']

columns = ["s5", "s1", "bp"]

returns ['s5', 's1']

Hints

solution.py

1from sklearn.datasets import load_diabetes
2from sklearn.ensemble import RandomForestRegressor
3from sklearn.inspection import partial_dependence
4from sklearn.model_selection import train_test_split
5
6
7def steepest_and_flattest(columns):
8    pass
9