User Based Collaborative Filtering Using Bloom Filter with MapReduce

Recommender systems help to solve excess information problem. Collaborative filtering is the most extensively used methods for recommendation. CF produces high quality recommendations based on likings of society of similar users. Collaborative filtering is based on assumption that people with same tastes choose the same products. Collaborative filtering does not perform well for large systems and it also suffers from sparse data. This paper proposes a novel approach where user based CF uses Bloom filter to filter out redundant intermediate results and helps to get better output. The bloom filter is implemented in the MapReduce phase.