Contextual Dependent Click Bandit Algorithm for Web Recommendation

In recommendation systems, it has been an increasing emphasis on recommending potentially novel and interesting items in addition to currently confirmed attractive ones. In this paper, we propose a contextual bandit algorithm for web page recommendation in the dependent click model (DCM), which takes user and web page features into consideration and automatically balances between exploration and exploitation. In addition, unlike many previous contextual bandit algorithms which assume that the click through rate is a linear function of features, we enhance the representability by adopting the generalized linear models, which include both linear and logistic regressions and have exhibited stronger performance in many binary-reward applications. We prove an upper bound of O(d\sqrt(n)) on the regret of the proposed algorithm. Experiments are conducted on both synthetic and real-world data, and the results demonstrate significant advantages of our algorithm.