Research on Intelligent Recommendation Algorithm of Short Videos Based on Graph Neural Network

The rapid development of short video platforms has put forward higher requirements for the accuracy and personalization of content recommendation systems. In this paper, a short video recommendation algorithm based on Graph Neural Network (GNN) is studied, which improves the recommendation performance by fusing multimodal features such as video, audio, and text. The key technologies such as graph convolution neural network, graph attention network and graph pooling operator are analyzed, and a multimodal recommendation framework is constructed by combining self-supervised contrastive learning and local feature encoder to effectively deal with complex user-content interactions. In this paper, several algorithms are compared on TikTok and MovieLens datasets. The experimental results show that the SHL algorithm significantly improves the recommendation accuracy and user personalized satisfaction on TikTok and MovieLens datasets, which is generalizable.