Question on year feature on weekly leaderboard pricing dataset

Hi @alfarzan
Another doubt - in the instructions, you write that in the weekly leaderboards you use datasets that go from year 1 to year 5. That is clear but - practically - you always set the year feature to 5 for the “new portfolio” or in this case you would use in the first weeks more year = 1 and later year = 5?

Many thanks

Hi @sandrodjay1

The weekly leaderboards do not have any year 5 data. It’s always 1 - 4. Can you point me to where it says 5 so I can fix it?

About your second question, for the leaderboards your model is used to make predictions for all rows in the dataset which includes data from years 1 - 4.

Is that clearer?

thanks alfarzan - it doesn’t say it anywhere, it was just a “technicality”. Because, in reality, once you parametrise your portoflio on 1-4, you always get a new portfolio with year = 5 … so I was wondering that, when you were saying that you were taking the year 1, you were then “overriding” the year to 5 as if in the “FINAL” leaderboard or if you were leaving the original year. I understood then you are leaving the original one. Perfect - thanks!

1 Like