Similar Item recommendations using PredictionIO
This template uses two algorithms:
- Picking random items
- MLib ALS + Cosine similarity
The first algorithm is a "pseudo" algorithm that just takes random items from the list of given items. This is useful when there is no user-item data at all. Once more and more data is available, the recommendations provided by collaborative filtering (Algo #2) will also start influencing the results, at which point Algo #1 can be either removed or its impact on recommendations reduced in case some element of randomness is still desired.
- python data/import_eventserver.py --access_key <>
- pio build
- pio train
- python data/send_query.py