Nowadays, E-commerce websites and recommendation systems are so common in our lives. From collected data in E-commerce websites, we found there are some repeated products in customers' buying history in different period. These consuming product will be purchased once again after customers run out of these products. Therefore, our recommendation system will focus on consuming product. The approach we used differs from previous recommendation systems by taking into consideration time series and repeated purchases. We think there is a strong relationship between time and consuming products according to users' behavior of purchasing repeatedly. On the other hand, we also try to use structured learning, which can add constraint on different users to get products which customers much more prefer. That is, it exists ranking concept in recommendation system, and we try to use structured learning to solve it. In this paper, we also design user features, item features, user-item features, and time-related features. Data collected is used with state-of-the-art Machine Learning algorithms. We establish a prediction model to test data split by time. The experiments shows that the top 5 F-measures are 66.38% higher than the previous study  done last year, and we thus providing an effective recommendation system.