💰 Must-Read: Use "predict_proba()" instead ...

Predictive Insights Youth Income Prediction Challenge

Helping South Africa

R10 000 ZAR

Challenge completed ~2 years ago

Skills you will learn

Prediction

Job Opportunity

637 joined

257 active

Info Data Chat Leaderboard

Start

Jun 08, 23

Oct 01, 23

Reveal

Oct 01, 23

University of Yaoundé I

Use "predict_proba()" instead of "predict"

Help · 4 Jul 2023, 17:55 · 9

predict_proba(X_test)[:, 1] is more adequat!

Discussion 9 answers

eskayML

Very helpful , especially for those who've made lots of submissions and are still unable to climb into 0.8

4 Jul 2023, 21:41

Upvotes 1

University of Yaoundé I

I thought the same thing since many people use the staternotebook as their base.

replied to eskayML4 Jul 2023, 22:06

Upvotes 2

DecScientist

I would have wanted to actually participate in this challenge because it is exciting! But I am in another one that's more to my personal liking, predicting the South African CPI. ButI agree with ff.

5 Jul 2023, 05:43

Upvotes 0

cephars

Free Lance

You can also participate,Iam also in RMB challenge

replied to DecScientist5 Jul 2023, 05:50

Upvotes 0

DecScientist

Will try. I just happen to have multiple use cases though, many of which to do with my employer, especially time series clustering and time series demand forecasting.

5 Jul 2023, 05:52

Upvotes 0

Satti_Tareq

I tried to apply your advice but it kept throwing errors, I googled the problem many times but the given solutions do not work.
ValueError: Unknown label type: 'continuous'

this is the message. Itried to label encode target as all solution does but it failed, I am boosting two lightgbm models.any help would be appreciated.

6 Jul 2023, 06:06

Upvotes 1

University of Yaoundé I

val_preds = model.predict_proba(X_val)[:, 1]

fpr, tpr, _ = roc_curve(y_val, val_preds)

score = auc(fpr, tpr)

I strongly advise you to study the "auc" metric so that you can better improve your score in the competition.

replied to Satti_Tareq6 Jul 2023, 07:33

Upvotes 1

Satti_Tareq

Thank you very much!

replied to ff6 Jul 2023, 09:03

Upvotes 1

diph

In R,

predictions <- predict(model, df_test_dummy, type = "prob")

# Then keep the second column

predictions <- predictions[2]

6 Jul 2023, 19:28

Upvotes 0

Join the largest network for
data scientists and AI builders

About FAQs

Status