Zindi User Behaviour Birthday Challenge
Helping Africa
$3 000 USD
Completed (~4 years ago)
Prediction
871 joined
174 active
Start
Sep 24, 21
Close
Jan 23, 22
Reveal
Jan 23, 22
'%3e%3ccircle r='20' fill='%23008'/%3e%3ccircle r='17.5' fill='%23fff'/%3e%3ccircle r='3.5' fill='%23008'/%3e%3cg id='d'%3e%3cg id='c'%3e%3cg id='b'%3e%3cg id='a' fill='%23008'%3e%3ccircle r='.875' transform='rotate(7.5 -8.75 133.5)'/%3e%3cpath d='M0 17.5.6 7 0 2l-.6 5L0 17.5z'/%3e%3c/g%3e%3cuse xlink:href='%23a' transform='rotate(15)'/%3e%3c/g%3e%3cuse xlink:href='%23b' transform='rotate(30)'/%3e%3c/g%3e%3cuse xlink:href='%23c' transform='rotate(60)'/%3e%3c/g%3e%3cuse xlink:href='%23d' transform='rotate(120)'/%3e%3cuse xlink:href='%23d' transform='rotate(-120)'/%3e%3c/g%3e%3c/svg%3e){kind=small}
What is the secret to 0.90 ?
8 Jan 2022, 16:21 · 6
Having a hard time breaching that 0.90.
CV LB both now stuck at 0.89 :(
Wow, trying all CV methods, I still can not get consistent CV/LB score....
How did you made it to 0.89? Only by FE and a realiable CV method?
Yes lot of Engineered features and making sure to avoid all sorts of Target leakage.
Training 10 kfold(shuffle=True) with separate tuned LGBM and sampled dataset for each month that needs predicting is what got me over 0.9.
Thanks Daniel, I will also test it out with my engineered features :). Till now, I have been training with 5 folds only.
Update: Disappointed with the shuffled 10fold performance, didn't help me any boost
Ahh, but are you using one model per target? There really is a ton of things we can try in this comp bc of how different the users are and how much legroom they give us with the datasets. As a data science newbie, I found this to be a pretty awesome comp to explore and keep experimenting. Still 14 days to go!