고객리뷰로부터제품/서비스하위기능의 우선순위를결정하기위한 BERT기반의 세밀한 IPA방법론

Abstract

The importance-performance analysis (IPA) is a widely-used method for prioritiz- ing target of improvement in products/services. IPA comprises three essential compo- nents. First, the analysis target corresponds to the feature of product/service. Second, performance evaluation pertains to the customer’s satisfaction associated with prod- uct/service. Third, the importance corresponds to customer expectations or expected value for the delivered product/service. Existing studies on the analysis of online customer reviews including IPA have made valuable contributions. However, they had limitations. The level of analysis targets in these studies was not sufficiently fine-grained. Although customers express specific features of a product/service, existing IPA studies have often overlooked or failed to distinguish these various facets of features. Consequently, these studies failed to identify detailed targets, leading to an inability to propose clear improvement pri- orities. To address this limitation, this dissertation proposes the sub-feature level IPA approach to suggest more detailed improvement target. Additionally, this proposed sub-feature level IPA approach is extended to trend and competitor analysis to de- rive more informative improvement priorities. To demonstrate the applicability of the developed approach, a case study of a television (TV) product was conducted. The outcomes of the study revealed several sub-features that were identified from four tar- get TV features. Additionally, the improvement priorities of the identified sub-features were determined from both the competitor and trend perspectives. This proposed ap- proach contributes to providing more specific and clear improvement targets, as well as fine-grained improvement priorities, from the perspective of the business. Keywords: Importance-performance analysis (IPA), product/service sub-feature, Performance, Importance, Competitor analysis, Trend analysis, Improvement target identification, Improvement priority, Bidirectional Encoder Representations from Trans- formers (BERT), Clustering method, Term frequency-inverse document frequency (TF-IDF), Contextual word map, Contextual meaning, Online customer review, Tele- vision, Sentiment analysis, Integrated GAP-performance difference (PD)-importance difference (ID).