Aiming at the problem of the separation analysis of short video propagation process and user behavior response in mobile social platforms, a comprehensive analysis framework combining content characteristics, platform mechanism, social structure and behavior feedback was constructed. Based on 52,000 short video samples and corresponding user behavior logs collected from January 2024 to December 2024, multi-modal content representation, propagation network analysis, clustering identification and predictive modeling methods are used to jointly test the propagation trigger mechanism, diffusion evolution mechanism and user behavior transformation path. The results show that topic focus, emotion expression intensity, information density and multi-modal presentation quality form the trigger basis of propagation, and the recommendation strength, hot traffic allocation and social relationship strength significantly affect the speed, scope and persistence of diffusion. Users show obvious heterogeneity in the dimensions of click, stay, comment, forward, follow and repeat visit. Low-active users are more dependent on the accuracy of recommendation, and high-active users are more likely to be stimulated by social interaction. Experimental results show that the Accuracy, F1 and AUC of the constructed model on the user behavior recognition task reach 0.903, 0.895 and 0.941, respectively, and the RMSE and R2 on the propagation effect prediction task are 0.109 and 0.872, respectively. This study can provide quantitative basis for short video propagation mechanism explanation, user behavior modeling and platform content governance.
