A Data-Driven Approach for Mining Truck User Requirements from Online Reviews

Abstract

The accurate identification of user requirements is essential for successful product design. However, existing methods for extracting user needs from online reviews often face challenges due to fragmented content and implicit expressions. To address these limitations, this study proposes a comprehensive framework for mining truck user requirements from online reviews. The process begins with the collection of online reviews to form an initial dataset, which is manually filtered for relevance and processed using NLP techniques to construct a domain-specific lexicon. We leverage K-fold cross-validation and LSTM neural networks to optimize text classification accuracy. Subsequently, sentiment analysis and quantification are per-formed on both explicit and implicit sentence structures, enabling the development of a requirement prioritization model that integrates IPA, the KANO model, and the DEMATEL method to quantify inter-requirement relationships. The results demonstrate that our approach effectively transforms unstructured user feedback into actionable product design insights, provides a reliable solution for manufacturers to improve truck design and development. This research not only advances the field of user requirement mining but also provides a practical tool for data-driven decision-making in the automotive industry.