Zimnat Insurance Recommendation Challenge
 scale(.21174)'%3e%3cpath d='m52.487 177.086 13.018 11.852 11.762-11.762 11.896 12.266 12.846-11.358 11.827 12.43 13.339-11.013 10.897 12.728 13.27-10.145 13.947 12.319'/%3e%3ccircle cx='89.425' cy='154.584' r='5.167'/%3e%3ccircle cx='111.237' cy='156.047' r='5.167'/%3e%3ccircle cx='30.05' cy='16.5' r='4'/%3e%3c/g%3e%3c/svg%3e)
Helping Zimbabwe
$5 000 USD
Challenge completed ~5 years ago
Prediction
Collaborative Filtering
1763 joined
612 active
Start
Jul 01, 20
Close
Sep 13, 20
Reveal
Sep 13, 20
'%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}
which is best algorithim for this data
Data · 12 Sep 2020, 09:42 · 12
can anyone suggest me which one was the best algorthim works for this data
So far Catboost.
Catboost and Feed forward neural networks get you in the 0.027-0.028 range that is for me this was the case. Eventually Catboost trumped it. Make sure to run any of the NN algos on Google colab to use its higher GPU bandwidth. Otherwise it is very very taxing.
You can also use kaggke privit notebook with GPU, it works well. The data is available there.
LightGBM ended up working the best for me, but it required a lot more tweaking than Catboost did.
Can you share your code. Would like to see your approach and learn from it.
5-fold LGB gave me the best result
Lightgbm in dart mode had the best results according to cross validation. However, catboost had the best results according to public and private leaderboard. I should have put more weight to catboost in my ensemble to score higher. At first, I thought that my catboost model were overfitting to public leaderboard, but it wasn't the case.
could your share please your params? for me catboost and xgb perform worse.
@lcfstat, just asking-how do you get to see how your model is peforming from the private leaderboard?
@ssshch { 'boosting_type': 'dart', "max_depth":-1, "num_leaves":32, 'learning_rate': 0.1,"min_child_samples": 20, 'feature_fraction': 0.8,"bagging_freq":1,'bagging_fraction': 0.9,"lambda_l1":1,"lambda_l2":1}
However, I think you should be carefull with the number of iterations because dart mode does not support early stopping rounds.
@elijah-a-w Once the competision has finished, all participants can see their private leaderboard scores in the submissions menu.
My approach was sum of lgbm multiclass and binary problems.
one hot of categorical features, tfidf and tuning lambda_l2 increased my score.(+/- 40 place on public and privat)