Prediction with Confidence in Item Based Collaborative Filtering

Recommender systems can be viewed as prediction systems where we can predict the ratings which represent users' interest in the corresponding item. Typically, items having the highest predicted ratings will be recommended to the users. But users do not know how certain these predictions are. Therefore, it is important to associate a confidence measure to the predictions which tells users how certain the system is in making the predictions. Many different approaches have been proposed to estimate confidence of predictions made by recommender systems. But none of them provide guarantee on the error rate of these predictions. Conformal Prediction is a framework that produces predictions with a guaranteed error rate. In this paper, we propose a conformal prediction algorithm with item-based collaborative filtering as the underlying algorithm which is a simple and widely used algorithm in commercial applications. We propose different nonconformity measures and empirically determine the best nonconformity measure. We empirically prove validity and efficiency of proposed algorithm. Experimental results demonstrate that the predictive performance of conformal prediction algorithm is very close to its underlying algorithm with little uncertainty along with the measures of confidence and credibility.