Radu

Hi Stephen. I have indeed managed to make it confusing, my apologies. The booking/click are just 2 features out of roughly 50. These contain details such as location, review scores, prices, etc. so the ranker would learn based on these features. The dataset is almost exactly the same as this one from Kaggle. However, there is no explicit target value, just the task of ranking hotels the best. The scoring thing I mentioned, so assigning a 7 if booked, 1 if clicked, 0 if nothing, is how my model will...

Hi Stephen. I have indeed managed to make it confusing, my apologies. The booking/click are just 2 features out of roughly 50. These contain details such as location, review scores, prices, etc. so the ranker would learn based on these features. The dataset is almost exactly the same as this one from Kaggle.

Hi Stephen. I have indeed managed to make it confusing, my apologies. The booking/click are just 2 features out of roughly 50. These other features contain details such as location, review scores, prices, etc. so the ranker would learn based on these features. The dataset is almost exactly the same as this one from Kaggle.

I'd like to use LambdaMART to rank the hotels from a bunch of searches. As a metric, I want to use nDCG@k with k the max number of entries in a search. The data I have has details mentioning whether the hotel was booked, clicked, or nothing and I want to assign a high importance to booking. For example, when calculating the score in nDCG, I'd like the book to be scored a 7, clicked a 1, and nothing a 0. If I set the target with values of either 7, 1, or 0 for each line of a search, will the nDCG...