search

PredictionIO / template-scala-parallel-universal-recommendation

PredictiionIO Template for Universal Recommender
111 stars 48 forks source link

Realtime Events consideration #56

Open hariyada opened 8 years ago

hariyada commented 8 years ago

Hi ,

    From my Application i am inserting events to event server in real time. If i query for recommendations
    the recommendation algorithm is not considering the new events that are uploaded to event server in    real time. Every time i end up in training and deploying the model explicitly for Better recommendations.

so my question is Is there a way that i can avoid the step of "pio train" and "pio deploy" for uptodate recommendations.

Regards, Hari

pferrel commented 8 years ago

This is not the root repo for the Universal Recommender. Get better response to questions by joining the Google Group here: https://groups.google.com/forum/#!forum/actionml-user

The Universal Recommender uses real time user input for recs queries and when you re-train the models is auto-deployed, so you may deploy only once ever. Train happens in the background as needed. Generally you re-train proportional to how often you add new items. The only restriction is that ant recommended item must be in the live model.