In recommender systems, user interaction will play an important role. During the prior analysis and studies the algorithmic recommender systems pointed the modeling approaches and enhancement for the future. The basis for designing and training the recommendation algorithms are formed by the feedback implicitly given the user and user ratings provided explicitly on the recommended items. But the user interactions in the web applications are ideal for the models provided in prior. To identify

more »

... address the problem, an online learning framework was built. The main focus of this paper is to provide an approach to interpret the users' actions. Our experiments on the large-scale data from a commercial web recommender system demonstrate significant improvement in terms of a precision metric over the baseline model that does not incorporate user action interpretation. The efficacy of this new algorithm is also proved by the online test results on real user traffic.