#!/usr/bin/env python3
# Source: q1-ml-fundamentals.html
import numpy as np
rng = np.random.default_rng(7)
X = rng.normal(size=(360, 3))
y = (X[:, 0] + 0.3 * rng.normal(size=360) > 0).astype(int)
user_id = np.repeat(np.arange(20), 18)
# Rows also represent time order in this synthetic example.
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.impute import SimpleImputer
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import GroupKFold, TimeSeriesSplit, cross_val_score

# Every fitted step lives inside the pipeline, so it is refit on each training fold.
pipe = make_pipeline(SimpleImputer(strategy="median"), StandardScaler(),
                     LogisticRegression(max_iter=1000))

# One user never appears in both train and validation.
scores = cross_val_score(pipe, X, y, groups=user_id, cv=GroupKFold(n_splits=5),
                         scoring="average_precision")

# Rows sorted by time. gap= leaves an embargo between train and validation.
ts_scores = cross_val_score(pipe, X, y, cv=TimeSeriesSplit(n_splits=5, gap=7))
print(scores.mean(), ts_scores.mean())
