Time:2019.04.04, 10:00 AM

Location:Room 615, Main Building

KeynoteSpeaker:Dr.Li Yang

Bio:Dr. LiYang is a professor of Cheung Kong Graduate School of Business, teaching EMBA,EE, FMBA, MBA, etc. Dr. Li Yang has provided marketing consulting for companiessuch as Tencent, Baidu, Yonghui Supermarket and Haier, and holds US patents formedical image processing.

Abstract:In this paperwe develop an ovel covariate-guided heterogeneous supervised topic model thatuses product covariates, user ratings and product tags to succinctly characterizeproducts in terms of latent topics, and specifies consumer preferences viathese topics. Recommendation contexts also generate big data problems stemmingfrom data volume, variety and veracity, as in our setting that includes massivetextual and numerical data. We therefore develop a novel Stochastic VariationalBayesian (SVB) framework to achieve fast, scalable and accurate estimation in suchbig data settings, and apply it to a Movie Lens dataset of movie ratings andsemantic tags. We show that our model yields interesting insights about moviepreferences and predicts much better than a benchmark model that only usesproduct covariates. We showcase how our model can be used for targeting recommendationsto particular users and illustrate its use in generating personalized searchrankings of relevant products.