Let the Held-Back Rows Decide

You have one set of points and several degrees to choose from. The honest way to choose is to hold some points back and see which degree does best on them.

Write pick_degree(xs, ys, degrees, seed).

Split the points with train_test_split(X, ys, test_size=0.3, random_state=seed), where X is xs as a one-column table.
For each degree in degrees, fit make_pipeline(PolynomialFeatures(degree), LinearRegression()) on the training part and score it on both parts.
Pick the degree with the highest test score, comparing scores rounded to 3 places. If two degrees tie, pick the smaller degree.

Return a dict:

"degree": the chosen degree, as an int.
"train": its training score, rounded to 3 decimal places.
"test": its test score, rounded to 3 decimal places.

Examples

xs = [0.34, 0.38, 0.45, 0.64, 0.95, 1.14, 1.56, 1.72, 1.73, 1.92, 2.07, 2.33, 2.35, 2.59, 2.94, 2.95, 3.21, 3.83]

ys = [12.05, 11.77, 12.18, 13.6, 14.29, 14.34, 14.93, 15.16, 15.71, 14.59, 14.29, 14.03, 13.2, 12.5, 11.35, 11.18, 9.69, 7.09]

degrees = [1, 2, 3]

seed = 0

returns {'degree': 3, 'train': 0.984, 'test': 0.944}

xs = [0.34, 0.38, 0.45, 0.64, 0.95, 1.14, 1.56, 1.72, 1.73, 1.92, 2.07, 2.33, 2.35, 2.59, 2.94, 2.95, 3.21, 3.83]

ys = [12.05, 11.77, 12.18, 13.6, 14.29, 14.34, 14.93, 15.16, 15.71, 14.59, 14.29, 14.03, 13.2, 12.5, 11.35, 11.18, 9.69, 7.09]

degrees = [1, 2]

seed = 1

returns {'degree': 2, 'train': 0.956, 'test': 0.914}

Hints

1import numpy as np

2from sklearn.linear_model import LinearRegression

3from sklearn.model_selection import train_test_split

4from sklearn.pipeline import make_pipeline

5from sklearn.preprocessing import PolynomialFeatures

8def pick_degree(xs, ys, degrees, seed):

9 pass