🏥 Challenge Chat: First Place Solution

Basic Needs Basic Rights Kenya - Tech4MentalHealth

Helping Kenya

$4 200 USD

Completed (over 6 years ago)

Skills you will learn

Classification

1257 joined

499 active

Info Data Chat Leaderboard

Start

26 Apr 20

05 Jul 20

Reveal

05 Jul 20

Klai

First Place Solution

Connect · 28 Jul 2020, 10:57 · 13

Congratulations to all the winners

our model is an ensembling of 3 roberta base with different params !

also we did some functions for data augmentation which help our model to fit Drugs and Suicide Tweets !

github link: https://github.com/Klaimohelmi/Basic-Needs-Basic-Rights-Kenya---Tech4MentalHealth/blob/master/final_kernel_rank3.ipynb

Discussion 13 answers

Gozie

28 Jul 2020, 11:06

Thanks Klai for sharing.

Congratulations!

Upvotes 0

Klai

28 Jul 2020, 11:14replied to Gozie

you're welcome

Upvotes 0

Lalit

28 Jul 2020, 11:36 (edited less than a minute later)

Congratulations! Klai and Thank you for sharing your work.

0.32*sub1[e]+0.68*sub2[e] OR sub7['Drugs']*0.8+0.2*sub5['Drugs']

would you please explain, how you guys calculate these coefficients(0.32,0.68)?

Upvotes 0

Hectogram

28 Jul 2020, 11:47 (edited less than a minute later)replied to Lalit

We assign weight of each model based on CV score. So one will give more weight on a model which "perform well". (Just a reminder, if you are choosing weights based on Public LB scores, then there is a chance that they will badly overfit.)

Upvotes 0

Lalit

28 Jul 2020, 12:12 (edited 1 minute later)replied to Hectogram

CV score? where did you calculate that in your code? if you don't mind, could you share any blog or steps to calculate this or how to assign these coefficients?

Upvotes 0

Hectogram

28 Jul 2020, 13:56replied to Lalit

Look at this blog.

Upvotes 0

Lalit

28 Jul 2020, 14:21replied to Hectogram

Thank you this is very helpful.

Upvotes 0

GBK

28 Jul 2020, 11:45

@Klai Congratulations!