# 7th Place Solution

Hi all
Thanks to the organizers for setting up the competition and many congratulations to the winners!
It was a great opportunity to experiment with ML techniques on what was a vaguely familiar type of dataset. I looked to learn from the public notebooks of the tabular playground series and started to experiment with AutoML libraries – for the submission I used h2o autoML as it is one of packages installed by kaggle.
If you are interested, I have tidied up my submission notebook and you can find it here:
[](https://www.kaggle.com/bernhardklinger/actuarial-loss-prediction-automl)
[https://www.kaggle.com/bernhardklinger/actuarial-loss-prediction-automl](https://www.kaggle.com/bernhardklinger/actuarial-loss-prediction-automl)
As highlighted in the discussion, the outliers in the dataset have a significant impact on training and validation.  I considered removing the outliers but left them in for the submission and fitted the models to predict the log of the target which seemed to improve the LB score.
Thanks again to all of you who contributed and shared ideas!