continuous training

User 386 | 7/9/2014, 5:49:39 PM

I would like to create a recommendation system based on the web pages people view and recommend them other pages they might like. For that matter, lets say I continuously have log files being written (to s3 or whatever other storage..lets say a local storage for this discussion) which include all the data rows.

How can I read multiple files at once into an SFrame?

I need to train a model and also give recommendations. how should I train my model (lets say, once a day) with all the new data that had arrived during the previous day, and give recommendations for the real time arriving data? is it possible to take and existing saved model, and at the same time train it with additional new data AND use it for creating recommendations?

any documentations for this subjects?

No Comments