I would like to use slope one as item reccomender. The problem is that I have groups of items that are not correlated each other.It seems to me that there is no way to tell Mahout to use the diff storage just for a group of products.I want to achieve this because the groups have an average of 100 items and I prefer to not create a mongoDbDiffStorage from scratch. Does the rescorer tell what differences should be computed in order to avoid to store useless data?
Thanks
You mean that you know that you want to consider some items completely unrelated? No, you'd have to write your own DiffStorage
for that.
Or, I might suggest that if these item sets are completely unrelated, then what you really have is a recommender problem for each group of items, and can use a Recommender
for each subset of data. This would probably be easier and more efficient in many ways.