USER BEHAVIOR PREDICTION BY FDNN MODEL ON SOCIAL NETWORKS
LE CHANG
Qianjiang College, Hangzhou Normal University, Hangzhou, Zhejiang, 310018, China.
LIDONG WANG *
Qianjiang College, Hangzhou Normal University, Hangzhou, Zhejiang, 310018, China.
KANG AN
Qianjiang College, Hangzhou Normal University, Hangzhou, Zhejiang, 310018, China.
*Author to whom correspondence should be addressed.
Abstract
Following the development of social networks, understanding user behavior can help some tasks in user recommendation, information dissemination, e-commerce recommendation and etc. Due to the non-availability of user data, it is still remain challenge to predict user behavior with high accuracy. In this paper, we propose a deep neural network FDNN (Fusion Deep Neural Network) based on BLSTM to predict two kinds of user behaviors simultaneously, such as comment and retweeting. Our model utilizes BLSTM to obtain high quality embedding space for user history and the new query post. We design a concatenation layer, a hidden layer and an output layer for behavior prediction. The experimental results on Sina dataset show that the proposed method could achieve better results than other traditional methods.
Keywords: User behavior prediction, BLSTM, deep neural network, social network