PhD defence H. (Hong) Deng

Below is a brief summary of the dissertation: 

This dissertation focuses on advanced methods to improve real-time personalization in marketing.

Chapter 2 proposes a novel real-time personalization engine to help find the optimal offer to provide to specific customers, thereby enabling effective customization in E-commerce. Yet, it remains challenging to optimize an offer strategy in real time, especially in a dynamic environment where the set of available offers varies over time.
We provide an easy-to-implement personalization engine to quickly learn and serve optimal context-dependent offers in situations where the offer set may change over time.

Chapter 3 addresses the challenge of using high-dimensional data that describes the context in real-time personalization. Personalization strategies in marketing often build on a large set of customer-specific and other contextual features. Contextual multi-armed bandit algorithms offer a principled framework for marketers to conduct sequential experimentation efficiently and to learn optimal personalized actions. However, conventional bandit algorithms are not well-suited for settings with high-dimensional features.
We propose a generic approach that adaptively learns the model specification to support the exploration process and effectively resolve the uncertainty in feature importance.

Chapter 4 introduces a method to achieve optimal targeting when the underlying reward patterns are time-varying due to seasonality or changes in consumer preferences. An initially successful campaign may later have adverse effects due to factors such as changes in competitors’ strategies or seasonality. Shifts in these relations may affect the optimal personalized actions and their profitability. We develop a contextual bandit algorithm with breakpoint detection to accommodate such non-stationary reward patterns.