Research on Collaborative Filtering Recommendation Method Based on Context and User Credibility

In the traditional collaborative filtering recommendation, similarity measurement methods only consider the user rating and the credibility of user rating is not taken into account, user's contexts are considered inadequate in the mobile environment, and the scalability problem exists in the recommendation system. A parallel collaborative filtering model based on user context and credibility is proposed. This method firstly evaluates user rating credit degree. Secondly, the method builds the context vector of the user, calculates the context similarity between the target user and other users, and searches for similar nearest neighbors for the target user based on trust and context, and finally implements the parallel recommendation on the cloud computing Mapreduce. Experimental results show that this method achieved lower error values of MAE than the traditional recommendation method and higher recommendation accuracy, and effectively improved the performance of the recommendation system. This method could be applied in the contextual recommendation oriented the big data.