As a new integration of the digital and real economies， digital cultural tourism presents fresh opportunities for the tourism industry. A key challenge is how to effectively leverage digital technologies to empower the transformation and upgrading of tourism enterprises based on evolving visitor needs. Quality Function Deployment （QFD） has proven to be an effective tool for translating Customer Requirements （CRs） into improvements in Technical Characteristics （TCs）. However， methodological limitations and modeling gaps remain across QFD’s core stages. In the requirement identification stage， traditional static approaches based on surveys and interviews fail to address the dynamic nature of visitor preferences； although recent applications of text mining and NLP improve automation， the mapping of extracted needs to structured design inputs remains inadequate. In the requirement weighting stage， QFD often applies the Kano model and sentiment analysis to categorize and assign weights to CRs. While this enhances objectivity， there is still no unified standard for weight modeling or classification boundary control. In the technical prioritization stage， although group decision-making and flexible modeling methods have been introduced to handle expert divergences， most studies rely on fixed consistency thresholds， lacking mechanisms and empirical validation for real-world collaborative settings.
To address these challenges， a QFD approach is proposed based on an improved Kano model and tolerance-driven group decision-making. In the requirement identification and weighting phases， customer needs are first extracted from online reviews using LDA and Word2Vec models， with a manually constructed dictionary. Then， the Kano model is enhanced by incorporating sentiment analysis and TF-IDF principles to categorize needs and assign quantitative weights. Initial weights are dynamically adjusted using prospect theory based on the enterprise’s development stage. In the technical prioritization phase， the classical group decision paradigm is refined by introducing a tolerance-based expert acceptance mechanism to replace fixed consistency thresholds. This yields expert weights and aggregated relation matrices， ultimately producing a consensus-based ranking of technical characteristics. A case study involving a national museum's digital tourism service design validates the feasibility and effectiveness of the proposed method.
This study makes the following key contributions： （1） a quantitative improvement to the Kano model that eliminates dependence on traditional Kano surveys； （2） a two-stage weight adjustment mechanism grounded in prospect theory to prevent excessive initial weight distortion； and （3） a tolerance calibration mechanism driven by expert acceptance rates， replacing theoretical thresholds and promoting a data-driven， progressively inclusive group decision process—offering new tools for QFD and similar application scenarios.
