Multi-Perspective Recommender Systems

Recommender systems play a critical role in online business as relevant personalized recommendations can have huge impact on users' consumption and purchase decisions. With deeper involvement of such automatic decision support in various specific contexts and tasks, concerns also arise regarding appropriate application and adaptation of the systems to generate high-quality recommendations. This stream of research aims at addressing some of the existing challenges in this field by providing multi-perspective solutions.

• From Head to Long Tail: Efficient and Flexible Recommendation Using Cosine Patterns. Yaqiong Wang, Junjie Wu, Zhiang Wu, Gediminas Adomavicius. Under Review at IEEE TKDE [SSRN]

• Enhancing Machine Learning with Domain Expertise: Effect of Stylists’ Outfit Composition Expertise on Apparel Recommender Systems. Gediminas Adomavicius, Shawn Curley, Yaqiong Wang. Work in Progress

• Cosine Regularized Matrix Factorization for Long Tail Recommendation. Gediminas Adomavicius, Xuan Bi, Yaqiong Wang. Work in Progress

Complementing Individual Predictions for Intelligent Decision Support

Applying data-driven decision support in real-world contexts is complex in that the automated predictions are often imperfect due to noisy, limited data or simplified mathematical or statistical assumptions. The imperfection of predictive models poses challenges in decision-making facilitation, especially in highly risk-sensitive domains like pharmaceutical research, medical diagnosis, or financial markets. This stream of research advocates multi-dimensional evaluation of data-driven predictive modeling solutions for the purpose of their intelligent applications and make methodological contributions to advance predictive-analytics-based decision support.

• Improving Reliability Estimation for Individual Numeric Predictions: A Machine Learning Approach. Gediminas Adomavicius, Yaqiong Wang. Accepted at INFORMS Journal on Computing, 2020. [Final Version]

Retailer Channel Management in the Digital Age

The rapid development of the Internet technologies and their applications to the retailing industry has made the online channel an imperative consumer touchpoint. In practice, retailers have experimented with multi-channel strategies by having online storefronts to supplement their existing pool of offline stores. This set of works combine different research methods to study the economic value stemming from the synergy between online and offline channels, understand whetehr and how different channels would improve overall business performance and adapt data analytics to avoid malicious product-related information manipulation in the online channel.

• The Role of Physical Stores in the Digital Age: Quasi-Experimental Evidence from Product Level Analysis. Jason Chan, Yaqiong Wang, Kaiquan Xu, Xi Chen. Work in Progress [SSRN]

• hPSD: A Hybrid PU-Learning-Based Spammer Detection Model for Product Reviews. Zhiang Wu, Jie Cao, Yaqiong Wang, Youquan Wang, Lu Zhang, Junjie Wu. IEEE Trans. on Cybernetics, 50(4), 2010. [Journal]