With the rapid development of social commerce, how to push and diffuse marketing messages in online social network (OSN) more effectively has increasingly become a significant issue, which can result in benefits for enterprises, users and platforms. A fundamental solution to this issue is how to accurately and comprehensively model user interest. To resolve such a significant and challenging task, our study constructed a user interest graph represented by a hierarchical tree structure that covers a wide range of topics, from coarse-grained to fine-grained three-level interest topics, such as food, entertainment and shopping, with a total of 167 nodes. In addition, considering that a user’s interests are always changing over time, an exponential interest decay scheme is employed in this study. Finally, a series of experiments are conducted to evaluate the performance of the proposed model by comparing it with three benchmarks designed based on the proposed algorithm and two similar hierarchical user interest models. The experimental results demonstrate our model works well to predict user interests. This research will provide important basic technology and valuable decision support for precise and personalized social marketing practices.

Introduction

Interpersonal interactions – such as information transmission, emotional communication, business transactions – have been enhanced by emerging and diverse online social networks (OSN), such as Facebook, Twitter and Dianping. 1 Gradually, a virtual society emerged, which dominates the bulk of human digital footprints, including relationships and behaviors (Lazer et. 2009, Freeman 2004). In this context, how to utilize behaviors, such as recommending, reviewing, forwarding and sharing among users in the virtual society (Mislove 2009, Zhu 2013) to carry out effective marketing activities (Li and Shiu 2012), has become one of the most important issues in the social commerce revolution (Steven and Olivier 2010, Han et al. 2018). The simple and direct pattern of push-forward-diffusion marketing messages by leveraging asocial graph has been widely adopted in current practice and academic research on social marketing (Turban et al. 2015, Zhu et al. 2016).). Yet, because of inadequate consideration of user interests and preferences, this pattern can easily result in an uninterested user’s antipathy in marketing message diffusion in OSNs. Moreover, this simple pattern’s lower precision will undoubtedly increase the marketing cost of enterprises, while the effects and efficiency often remain unsatisfactory. Further, to improve user experience and enhance users’ stickiness, a growing number of social platforms have also begun to restrict the indiscriminate flooding with marketing messages. Therefore, the lack of precision and personalization in current social marketing practices has been a prominent problem that brings trouble to users, enterprises and platforms (Burchell et al. 2013). It is worth noting, though that there actually is a user interest graph in OSN besides the user social graph shown in Fig. 1 as an example.