Identification and analysis of driving factors for product evolution: A text data mining approach

Abstract

Products continuously evolve over time as a result of changes in various factors, including technological advancements, customer requirements, and material processes. The complexity and interconnections among these factors present significant challenges for their identification and description. Traditional studies primarily rely on inductive summarization, which often faces issues of subjectivity, uncertainty, and low reliability. This research presents a method combining the Bidirectional Encoder Representations from Transformers (BERT) model and Dynamic Topic Model (DTM) to analyze the driving factors of product evolution. First, the BERT model was employed to enhance the DTM model, and a text corpus related to product evolution was constructed to identify its driving factors. Then, similar algorithms and co-occurrence network analysis methods are applied to study the spatiotemporal evolution of these driving factors at different granular levels and their impact on designers' cognition. Finally, a case study on the evolution of automobiles is conducted to verify the effectiveness and applicability of the proposed model. The results indicate that incorporating the BERT model to enhance the DTM model improves semantic extraction from textual data. Moreover, significant interdependencies were identified among the driving factors, with their specific meanings progressively evolving towards domains such as human emotions, culture, and experiences. From a data mining perspective, this approach addresses the challenges of identifying product evolution driving factors, assisting designers and decision-makers in executing iterative product development more efficiently and scientifically.