I would index the products a user purchased as well as the number of times purchased, then I would take a user, search their bought products boosted by how many times purchased, against other users, have a facet for products and filter out the top bought products that are not on the users already purchased list. then you have a list of products purchased by users with the same buying habits as your user that they have not bought. And over time you can tune your original search with geographic info or age or other demographics that return more relatable users etc.
Fun mental project, would be fun to have a data set like amazon or wal mart or something to see if you start getting legit results. Could even see if you could start predicting their next purchase when you throw in things like time of year and recently purchased items, cross reference with the other users etc. with enough data I really start enjoying weaponizing solr, it can be quite entertaining as long as you have no morals with privacy or they clicked the little box allowing you to do anything you want. But that’s how Facebook and the likes make a lot of money, is by taking your friends and following them around the Internet and doing the above to place the exact ad for a bottle of wine that your best friend just bought a few weeks ago and brought to an event you were invited to. Gets addicting:) > On Dec 10, 2019, at 5:56 PM, Arnold Bronley <arnoldbron...@gmail.com> wrote: > > Hi, > > I have a Solr collection 'products' for different products that users > interact with. With MoreLikeThis, I can retrieve for a given product > another related product. Now, I want to create a Solr collection for users > such that I can use MoreLikeThis approach between users and products. Not > just that, I would also like to get relevant product for a user based on > some sort of collaborative filtering. What should be my indexing indexing > and collection creation strategy to tackle this problem in general?