Time and Location-Based Hybrid Recommendation System

With the development of the Internet, e-commerce industry rises rapidly. Online shopping becomes more and more convenient and fast. However it is very difficult for consumers to find satisfied commodity because of abundant and mixed commodity. Especially when people purchase the items which they are not familiar with or consume in a strange place. The study of the recommended system is to figure out those problems referred above. Traditional recommended methods use the similarity of user and item, in addition some people use the content of the item. This paper will convert recommended problem into classification problem and merge together the knowledge-based method. We also propose a flexibility and scalable feature selection approach. As we know, people's activeness and item's popularity change with time. So we abstract time context from time window and time decay in this paper. Though location-based recommendation becomes more and more important, geography information often sparse. In order to solve the problem we apply a new method to fill the missing location information and abstract spatial properties of users and items. The result of recommendation had been improved by using our methods.