Research on product iteration direction prediction method based on perceptual-cognitive data of Danmaku

To enhance user feedback precision and development agility in high-cycle industries, this study utilises danmaku comments as a real-time data source for iterative innovation. We propose a methodology for predicting product iteration directions grounded in a Time–Sentiment–Feature mapping model, addressing the latency and retrospective bias characteristic of traditional feedback channels. Using smartphones as the research object, a fine-grained dataset was constructed via element-specific video segmentation. Satisfaction across design features was quantified through a hybrid sentiment analysis approach integrating a customised danmaku lexicon with the XLNet model. By synthesising these metrics with TextRank-extracted keywords, prioritised iteration plans were generated. Comparative validation against two product generations supports the framework’s utility in capturing spontaneous, pre-reflective user concerns. While particularly effective for high-engagement consumer electronics, the findings suggest that addressing danmaku-derived pain points serves as a critical driver for informing successful product iterations.