Recommender System Based on Collaborative Filtering and word2vec for Email Platform

• 1、School of Information and Communication Engineering,Beijing University of Posts and Telecommunications,Beijing 100876

Abstract：With the development of Web2.0 technology , the information the user really want submerges in the large amounts of information, the so-called information overload problem. So recommendation systems are becoming more important in e-commerce and Internet companies, which helps companies better understand and analyze users, then recommends for users of goods and services according to their tastes and interests. E-mail has been a indispensable tool with the appearance of the Internet, it contains a lot of the user's personal interests and hobbies features. This paper presents a recommendation system based on e-mail platforms, and use word2vec algorithm and Collaborative Filtering in order to better address the cold start problems which is a common problem for recommendation system. We analyze user mail messages and user behavior logs for user modeling, then generating user interest matrix by the Chinese word segmentation algorithm. We use books as our database to recommend. We recommend books based on the user interest similarity matrix and feature matrix of books. Besides, in order to eliminate drift and enhance user interest robustness of the system, we added a feedback system according to the user's behavior. When a user clicks one thing, the system will get similar things and then recommend to the user. The similarity between books is stored in thing similarity matrix, which was calculated by content-based and collaborative filtering techniques. what's more, if the user has only a few mails, we use word2vec algorithm to solve the cold start problem by performing word similarity expansion.

Keywords： Artificial Intelligence；Data Mining Recommender System word2vec；Email Platform Interest Matrix