Predicting Restaurant Consumption Level through Social Media Footprints

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 3328–3338, Osaka, Japan, December 11-17 2016. Predicting Restaurant Consumption Level through Social Media Footprints Yang Xiao, Yuan Wang, Hangyu Mao, and Zhen Xiao School of Electronics Engineering and Computer Science, Peking University, China {xiaoyang, wangyuan, mhy, xiaozhen}@net.pku.edu.cn Abstract Accurate prediction of user attributes from social media is valuable for both social science analy- sis and consumer targeting. In this paper, we propose a systematic method to leverage user online social media content for predicting offline restaurant consumption level. We utilize the social login as a bridge and construct a dataset of 8,844 users who have been linked across Dianping (similar to Yelp) and Sina Weibo. More specifically, we construct consumption level ground truth based on user self-report spending. We build predictive models using both raw features and, espe- cially, latent features, such as topic distributions and celebrities clusters. The employed methods demonstrate that online social media content has strong predictive power for offline spending. Finally, combined with qualitative feature analysis, we present the differences in words usage, topic interests and following behavior between different consumption level groups. 1 Introduction Over the past decade, microblogging services like Twitter and Sina Weibo have built up a huge user base. For example, by the end of 2014, Sina Weibo has accumulated more than 500 million users, out of which 167 million are monthly active users. With the growing popularity of microblogging service, businesses are also looking for new opportunities on social media, e.g., identifying target consumers and marketing their products. Compared with traditional consumer targeting scenarios, social media has several factors in its favour. Firstly, traditional consumer targeting techniques are mainly based on users’ query logs or web access histories, and it is generally limited by session length (Dasgupta et al., 2012). While on social media, users have accumulated much more abundant traces, such as tweets, relationships and profiles. Secondly, according to a recent study (Nielson, 2012), 92% of consumers believe recommendations from friends over all other forms of advertising and 64% of salesperson believe word of mouth is the most effective way of marketing. Since social network links both friends and families, it becomes a natural platform for business to take advantage of word of mouth utility (Trusov et al., 2009). Demographic profile is the starting point of defining target consumers for marketing. Demographic includes multiple aspects such as simple attributes like gender and age, and more complicated attributes like income, personality and consumption level. Since a large number of users provide their profiles on social media, inferring user attributes such as gender (Ciot et al., 2013; Liu and Ruths, 2013; Rao et al., 2011), age (Al Zamal et al., 2012; Nguyen et al., 2013), political polarity (Volkova et al., 2014), or occupation(Preot¸iuc-Pietro et al., 2015a) from social media has already been widely studied. In this paper, we focus on the consumption level attribute prediction. Consumption behavior reflects one’s economic capacity and living standards (Brewer et al., 2012). …

Content truncated to 500 words