Improving Customized Recommendation Accuracy Including User's Expectancy in Bandwagon Phenomenon

摘要

With the number of users and items has grown rapidly in online, recommender systems have come to play an important role in terms of helping users to choose items. In the process of consumer decision making, user tend to follow other people's opinions as a way to reduce the risk of making bad decision which is known as bandwagon effect. So if users see an item with high popularity that means that the other people also like the item, they will have a look forward to the item which can be seen as users' expectancy. However, very few studies have considered the impact of the expectancy on the users' evaluation of items. In this paper, we not only define user's expectancy but also standardize it by considering different users have different rating standards. Then we have developed a recommendation system that considers the psychological concept of bandwagon effect. As a result, our mechanism outperforms the existing ALS (Alternating Least Square) model in improving the prediction accuracy on RMSE.