Reasearch on User Profile Based on User2vec

Personalized services for information overload are becoming more common with the arrival of the era of big data. Massive information also makes the Internet platform pay more attention to the accuracy and efficiency of personalized recommendations. The user’s profile is constructed to describe the user information of the relevant platform more accurately and build virtual user features online through user behavior preference information accumulated on the platform. In this paper we propose a new user mode named user2vec for personalized recommendation. The construction of user2vec relies on platform and extremely targeted. At the same time, user profile is dynamically changing and need to be constantly updated according to the data and date, therefore we define a new time decay function to track time changes. Dynamic description of user behavior and preference information through user vectorization combined with time decay function can provide reference information for the platform more effectively. Finally, we using a layered structure to build an overall user profile system. And the experiment adapts content-based recommendation algorithm to indirectly prove effectiveness of user profile model. After many sets of experiments proved, it can be found that the proposed algorithm is effective and has certain guiding significance.