The rapid development of new media technology has increased the complexity of predicting the communication effects of language and culture in Chinese international education. This paper combines the Necessary Condition Analysis (NCA) and Qualitative Comparative Analysis (QCA) methods to identify the necessary conditions and combinations of conditional factors that affect the dissemination effect of short videos on language and culture in Chinese international education. A temporal convolutional network (TCN) is constructed to realize the prediction of communication effects of short videos in the new media framework. A multilayer deep time convolutional extended residual network network structure (MDTCNet) is proposed to optimize the prediction accuracy with respect to the prediction lag of TCN. The results show that the condition of “content theme” simultaneously satisfies the efficiency measure d > 0.2, with a p-value of <0.05, and the consistency index is 0.848, close to 0.85, which is a necessary condition for the high-quality dissemination of short videos about language and culture in Chinese international education. The existence of three combinations of conditional factors has strong explanatory strength for the dissemination effect. The improved MDTCNet model propagates heat prediction error of no more than 0.1 with an R² score of 0.88 for its prediction. The value is closer to the real value. Using the MDTCNet model to process the short video related condition data can effectively improve the prediction accuracy of the dissemination effect of language and culture short videos in Chinese international education.
