Integrated Machine-Learning Algorithm for Identifying Segment-Level Key Drivers from Consumers’ Online Review Data

I propose a newly integrated machine-learning methodology to apply a classic model-based segmentation method to unstructured online review data. The proposed algorithm extracts an independent variables matrix from unstructured textual reviews by developing a set of text-mining algorithms and then identifies segment-level key drivers by applying a proposed Bayesian ordinal probit mixture regression with variable selection. With the proposed method, firms can focus on key drivers per each segment in their marketing activities (e.g., online banner advertising, search advertising); this method will help them systematically keep track of periodic patterns of segment-level key drivers. Using online data from a large review site for rating professors, I validate the extracted independent variables through multiple validation studies and then show heterogeneous key drivers for satisfaction across three derived segments. For the least satisfied segment, the proportion of reviewers is significantly higher from the Science, Technology, Engineering, and Mathematics education category.