User-User recommendation recipe?

Hi, it sounds like you might be interested in using the Amazon Personalize user segmentation recipes. These recipes generate segments of users based on item input data.

A possible application couId be: if you want to recommend other users to user A, your input item, could be the last item user A interacted with, or the last category of items user A interacted with. Amazon Personalize will recommend users that have affinity for those items or item attributes. As long as you have enough interaction data and item metadata, you can generate recommendations. Check the personalize cheat sheet.

Alternatively, the Amazon Personalize custom dataset groups are use-case agnostic: you can use any kind of "items" to recommend to your users, including other users! You can create your items dataset with user_ids and metadata. However, in this case, the interaction data would be users interacting with other users, not with items.

As long as you have enough interaction data (and ideally some user and "item" metadata) you can generate recommendations. Check the personalize cheat sheet.

For both recommendation types you can also apply business rules to further improve your recommendations through filters . For example you can recommend only those users that have the same interest or type of the original user, or only users that have a similar company size.