Content Recommendation System Based on Global Data

In this paper, we make a simple review on the existed recommendation systems. In order to recommend appealing commodity to user, we make some improvements to content-based recommendation system. During the user is browsing the webpage, the keywords and the number of times they appear will be recorded as the on-line data. And user's operation on keyboard and mouse will be counted as the off-line data. These two kinds of data are so we called user's Configuration Documentation of Global Data (CDGD). Using K-means clustering algorithm, we can initially divide CDGD into different types. Then we use TF-IDF method to further characterize each kind of CDGD. By the method above, we can finally get characterized result vectors from different kind of keywords in each kind of CDGD. When there comes a new CDGD, it will be easy to collect new characterized result vectors. We sorted the new vectors by calculating the scalar product between CDGD and the new one. The software design and future expectation are given in the end.