PhD Project: Weiqiang Lin

Mining Clicks and Social Media for E-Commerce Optimisation

[All PhD Projects]

Wyatt Weiqiang Lin works on understanding how click-streams on e-commerce platforms and social media can be used for optimising prediction and personalisation in the project “Mining Consumer Feedback and Designing Features for E-commerce Platform Optimisation”. Wyatt is placed within the Big Data team within China’s largest retail JD.com with full access to databases.

Pearson Correlation between social media activities (Post, Repost, and Comment) and specific e-commerce platform
activities (Search, Clickthrough and Order) for 33 vendors using their two year time series. Heat maps show low to medium
correlations between SMA and EPA pairs. From “Lin W., Saleiro P., Milic-Frayling N. and Ch’ng E. (2018) Social Media Brand Engagement as a Proxy for E-commerce Activities: A Case Study of Sina Weibo and JD. In Web Intelligence, IEEE/WIC/ACM International Conference.”

Modeling user behavior based on clickstream data from e- commerce platforms has been an essential means of purchase predictions, consumer targeting, and personalised advertisement. However, approaches to data analyses have been ad hoc, lacking a framework for systematic refinement of predictive methods and improvement of e-commerce platforms. In this paper, we take a system design view and consider the entire purchase process facilitated by the e-commerce platform. Starting with a commonly adopted consumer Purchase Decision Model (PDM), we develop a framework for analysing on-line shopping sessions in terms of user activities that are facilitated by the platform, from product browsing and comparison to searching for deals and providing product feedback. By constructing multi-action features, including frequent action sets and sequences, we capture consumers’ purchase decisions within the product sale cycles. This enables us to create highly accurate prediction models of shopping outcomes. Based purely on user actions, with no product specific information, user preferences or purchase history, our method and models serve as a baseline for further clickstream analysis. Through experiments with data from a large e-commerce platform, we demonstrate that our framework is extendible and effective in leveraging feature refinements to improve purchase predictions.

[All PhD Projects]