Utilizing and discovering hidden information: strategic usage of curiosity and predicting personal traits

Abstract

Data allows us to figure out the information about the customers without directly observing their behaviors. Service providers can easily accumulate and utilize information about the customers as digitalization progresses in various fields such as commerce or healthcare. However, problem-solving with a data-driven approach requires preprocessing of data and model application (e.g., economic or machine learning models) as information from raw data such as customers’ demographics (e.g., gender or age), purchase history, or app logs cannot solely provide practical implications. This dissertation comprises two essays that discover and utilize the hidden information through the data. The first essay examines customers’ responses to the advertising strategy which intentionally conceals the information about the product at the beginning of the ad, by conducting large-scale field experiments with a book retailer. The second essay detects Parkinson’s Disease patients in the early stage by using time-sequential diagnostic history data with the Long Short-Term Memory algorithm.