The dissemination model of intangible cultural heritage images on new media platforms based on machine learning

Abstract

This study investigates how intangible cultural heritage (ICH) videos circulate across heterogeneous new media environments and proposes a platform-sensitive prediction framework. We represent each video through jointly encoded textual, visual, and optional acoustic signals, and then introduce an adaptive fusion mechanism that adjusts content features to different platform contexts. On this basis, the model estimates click-through probability and engagement potential, producing a dissemination score for cross-platform ranking. Experiments indicate that the proposed framework improves category recognition, audience-response prediction, and recommendation performance relative to common baselines. Beyond overall accuracy gains, the results show that platform differences matter: identical ICH content can display distinct propagation patterns under different recommendation logics and user communities. The study therefore offers both a quantitative tool for dissemination analysis and a practical route for increasing the visibility of ICH content in digital communication.