mapie/predict-and-predict-interval
| Rule | predict-and-predict-interval |
|---|---|
| Pack | MAPIE |
| Severity | warning |
| Fix | none, edit by hand |
| Library versions | every version |
predict_interval already returns the point predictions as its first value, so calling predict too runs the model twice. Unpack both from one predict_interval call.
y_pred = mapie.predict(X_test)_, y_pis = mapie.predict_interval(X_test)y_pred, y_pis = mapie.predict_interval(X_test)What the rule reports
Section titled “What the rule reports”The rule reports predict on the same object and the same rows as a predict_interval or
predict_set in the same block.
How to fix it
Section titled “How to fix it”xpd has no automatic fix. The rewrite needs a new name and changes two statements. Make the change by hand.
predict_interval and predict_set return the point predictions as their first value. With a
forest of 50 trees on one thread, both forms returned the same arrays, bit for bit, and the pair
of calls took longer than the single call:
| Calls | At 2,000 rows | At 20,000 rows |
|---|---|---|
predict and predict_interval, against predict_interval alone |
1.99 times | 1.99 times |
predict and predict_set, against predict_set alone |
1.98 times | 1.97 times |
The two calls must be statements of the same block and name the same object and the same rows.
Message
Section titled “Message”xpd prints this message with the issue. A name that starts with $ stands for code from your file.
$M.predict_interval($X) already returns the point predictions as its first value, so this predict runs the model a second time. The pair measured 1.99 times as long as predict_interval alone. Write y_pred, y_intervals = $M.predict_interval($X), which gives the same arrays.Browse the other rules of the MAPIE pack.