Skip to content

Add simple DatetimeEncoder example with periodic encoding (#1629) #1834

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
TaobaOuathrani wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Add simple DatetimeEncoder example with periodic encoding (#1629) #1834

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
75 changes: 75 additions & 0 deletions examples/0030_datetime_encoder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -352,3 +352,78 @@
# features from a datetime column.
# Also check out the |TableVectorizer|, which automatically recognizes
# and transforms datetime columns by default.

"""
Simple example: understanding what DatetimeEncoder extracts
------------------------------------------------------------

This example demonstrates what the DatetimeEncoder does on a simple,
non-forecasting regression task.
Copy link
Member

@rcap107 rcap107 Jan 14, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please expand a bit this section explaining that for the sake of the example we are using a synthetic dataset, and a plot of what the target looks like.


We compare a naive baseline using a raw numeric timestamp with a model
that uses DatetimeEncoder with periodic encoding to extract meaningful
time-based features for a linear regression model.
"""

import numpy as np
import pandas as pd

from skrub import DatetimeEncoder, ApplyToCols
from sklearn.linear_model import LinearRegression
from sklearn.pipeline import make_pipeline
from sklearn.metrics import r2_score

# ---------------------------------------------------------------------
# Create a simple synthetic dataset
# ---------------------------------------------------------------------
rng = pd.date_range("2023-01-01", periods=300, freq="h")

X = pd.DataFrame({"date": rng})

# Target depends on hour of day and weekday (cyclic pattern)
y = (
10 * X["date"].dt.hour.isin([8, 9, 17, 18]).astype(int)
+ 5 * (X["date"].dt.weekday < 5).astype(int)
+ np.random.normal(0, 1, size=len(X))
)

# ---------------------------------------------------------------------
# Baseline model: naive numeric timestamp
# ---------------------------------------------------------------------
X_baseline = pd.DataFrame(
{"timestamp": X["date"].astype("int64") // 10**9}
)

model_baseline = LinearRegression()
model_baseline.fit(X_baseline, y)

y_pred_baseline = model_baseline.predict(X_baseline)
print("R² with naive timestamp:", r2_score(y, y_pred_baseline))

# ---------------------------------------------------------------------
Copy link
Member

@rcap107 rcap107 Jan 14, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There should also be a section with a DatetimeEncoder that does not include periodic features for the sake of the comparison.

Copy link
Member

@rcap107 rcap107 Jan 14, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Also please add a few lines explaining in words what is happening here as part of the narrative of the example.

# Model using DatetimeEncoder with periodic features
# ---------------------------------------------------------------------
model_datetime = make_pipeline(
ApplyToCols(
DatetimeEncoder(
add_weekday=True,
periodic_encoding="spline",
),
cols=["date"],
),
LinearRegression(),
)

model_datetime.fit(X, y)
y_pred_datetime = model_datetime.predict(X)

print("R² with DatetimeEncoder:", r2_score(y, y_pred_datetime))

# ---------------------------------------------------------------------
# Inspect generated datetime features
# ---------------------------------------------------------------------
apply = model_datetime.named_steps["applytocols"]
encoder = next(iter(apply.transformers_.values()))

print("Generated features:")
print(encoder.get_feature_names_out())
Copy link
Member

@rcap107 rcap107 Jan 14, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please add a conclusion summarizing briefly what is being done here

Loading