A collaborative recommendation algorithm based on ratings and trust

Trust is used for recommendation which can solve the cold start and the cheating rates problem in conventional recommender system, but it is difficult to build the trust network and the preference relations among users. A collaborative recommendation algorithm is proposed based on ratings and trust, and the correlation expressions and the flow of algorithm are also presented. The similarity weight is calculated by the rating value and the trust value. The algorithm couples the ratings with the trust to establish the similarity weight, and the predicted ratings produce the candidate set for the target users. Experimental results and comparisons with the traditional collaborative recommendation and the trust recommendation show that the proposed algorithm greatly improves the coverage with a tiny loss in accuracy. The rating coverage is much higher than that of the traditional collaborative recommendation and the trust recommendation by 3% and 32.1%, respectively, and much higher than the traditional collaborative recommendation and the trust recommendation on the user coverage by 8.2% and 15.1%, respectively. And a perfect balance between the accuracy and the coverage is obtained.