Compare forecasts on the same later days

Accept the window where the fitted model loses

Botie needs tomorrow's water demand before tomorrow arrives. A project report must preserve the available history, stored forecasts, baseline, and matching later outcomes.

BOTIE

The baseline row [10, 10] and model row [14, 16] face actual demand [10, 11]. What should the report preserve?

The inputs we are comparing

Rows of matched procedure forecasts

Trend example

10	12
12	14

Model loses

10	10
14	16

Later observed outcomes

Trend example

Model loses

Result for trend example

Choose an explanation to test

Only the earlier trend example where the model wonRefit using these target values before reporting their forecastsThe model's worse error and the same-date baseline comparison

A weekly rule faces an unusual middle day

This check uses the pieces from “Compare complete forecasting procedures on matched days”.

BOTIE

Botie compares two forecasting rules on the same three held-out days. The first row repeats a daily baseline; the second follows a fitted weekly pattern. Return each row's mean absolute error against the actual demand.

Rows of matched procedure forecasts: 15 15 15
10 20 10
Later observed outcomes: 012
124
212

What result should your machine produce for this case?

Use a number, text, or an array matching the requested result. For a matrix, put each row in brackets. Saving your prediction runs your current construction on these inputs and checks the earlier examples too.

Build a construction in the workspace above first. This case will test those pieces.

Which changes to a future target should leave earlier stored forecasts unchanged, and how would you test that?

Your notes for later. These stay with this journey.

Retain origins, predictions, actuals, baseline errors, and fitted state. Audit calendar gaps and delayed readings before forming lagged inputs.

Forecasts must be stored before targets are observed. This workshop evaluates supplied predictions; fitting happens in the project code.

Work through the Python challenges →

Continue the courseCount Seeds While Keeping the Image Assumptions Visible →

← Investigate a Sensor Before Calling It Faulty