商学院卓越学术讲坛

讲座题目：

Image Network and Interest Group – A Heterogeneous Network Embedding Approach for Analyzing Social Curation on Pinterest

主 讲 人：马力烨

讲座时间：2022年11月23日（周三）10:00

腾讯会议：会议号 982-499-568

主讲人介绍：

马力烨博士是马里兰大学史密斯商学院的终身教职副教授，主要讲授大数据与人工智能、数据科学等课程。他的研究利用统计计量以及机器学习等方法分析数字经济下的消费者行为及其与企业的互动，并据此帮助企业制定营销策略，马教授的多项研究成果发表于营销领域顶级期刊，如 Marketing Science、 Journal of Marketing Research、 Journal of Marketing、 Management Science 等。马教授目前还担任营销领域权威期刊International Journal of Research in Marketing 的 AE，顶级期刊 JMR、JM 的编委会成员。马教授的研究曾获得 Marketing Science Institute和 Wharton Customer Analytics Initiative 的研究支持，论文还曾入围 John D.C. Little 最佳论文奖。

讲座摘要：

Social curation platforms help consumers navigate through vast digital content online. Analyzing a large dataset from the popular image curation site Pinterest.com, this research aims to understand: (i) what users’ curation activities reveal about consumer preferences, content characteristics, and brand perceptions; (ii) how to assess the user-content match and predict curation actions; (iii) how well does social curation site facilitate information discovery.

We propose a novel approach with two components. First, we represent social curation using a heterogeneous information network. Images, users, and curation words are represented as nodes, while edges represent curation actions. Second, we leverage heterogeneous network embedding, a recently developed machine learning method, to map the network to lower-dimensional vectors for analysis while preserving its structural and semantic information.

Our proposed approach significantly outperforms prevailing benchmarks on predicting curation actions. It uncovers user interest groups and image clusters with distinct characteristics, characterizes the user-image matching, and generates insights into brand perceptions and opportunities. Furthermore, analysis shows that social curation activities account for 70% of the information content at the site, while algorithm bias is not evident. We are the first to study social curation using an information network approach, and our study provides ready-to-use tools for managers.