Modeling Dynamic Consumer Preferences in Product Attributes for Social Media-Based Product Improvement Planning

Abstract

Consumer preferences for products change over time. The changes in consumer preferences are not just related to the level of importance of the attribute but also dynamics in the attribute itself. Previous studies generally employed the pre-determine attribute to measure the level of importance of the attribute. However, the pre-defined attributes approach has a problem in that new consumer attributes are not captured, resulting in misleading product development. Therefore, this study proposes a method to identify dynamic attributes and the level of importance of the attribute for product development without pre-determined attributes and unsupervised. A case study on the airline industry demonstrates the proposed method. Attribute changes are assessed using latent Dirichlet allocation to identify topics related to product attributes and sentiment analysis to evaluate satisfaction levels. Furthermore, the association rule approach is used to label unsupervised product topics. A total of 222,959 tweets related to the airline industry from 2015 to 2018 were collected and analyzed to demonstrate changes in consumer preferences over time. The results indicate that the proposed method effectively explores the dynamic of product attributes and critical attributes and provides valuable strategies for developing product or service improvement. The proposed method has also been validated and is in line with the successful innovation strategies of an Indonesian airline.