Online recommender systems often deal with continuous, potentially fast andunbounded flows of data. Ensemble methods for recommender systems have beenused in the past in batch algorithms, however they have never been studied withincremental algorithms, that are capable of processing those data streams onthe fly. We propose online bagging, using an incremental matrix factorizationalgorithm for positive-only data streams. Using prequential evaluation, we showthat bagging is able to improve accuracy more than 35% over the baseline withsmall computational overhead.
展开▼
