As far as I know tree-based models are always on top of leaderboard in ML competitions. So tuning the top 3 ones is a good choice: xgboost, catboost, and lightgbm. I've just tried those with default settings and it put me on a descent position with no tuning.
Of course if you have enough time you can also try neural nets, with enough tuning and architecture design they can compete with gradient boosted
As far as I know tree-based models are always on top of leaderboard in ML competitions. So tuning the top 3 ones is a good choice: xgboost, catboost, and lightgbm. I've just tried those with default settings and it put me on a descent position with no tuning.
Of course if you have enough time you can also try neural nets, with enough tuning and architecture design they can compete with gradient boosted
Thank you @alka.
did u used along with multioutput regressor
No I tried the naive approach of fitting a single model per output. But I it is functionally equivalent to using sklearn multioutput Regressor wrapper