The current process of education informatization is deepening, and the demand for intelligent technology in the field of education is becoming more and more urgent. This paper proposes a deep composite recommendation model VAE-GAN-DCR based on variational autoencoder and generative adversarial network, and explores the effect of generative artificial intelligence in smart classroom. Methodologically, the model combines the decoder of VAE with the generator of GAN, improves the traditional VAE model by introducing a priori distributions that depend on item features, and optimizes the reconstruction error by using the feature transfer of the GAN discriminator to achieve accurate recommendation of educational resources. At the same time, the Williams Creative Tendency Measurement Scale is used to evaluate the teaching effect of students in Zhanjiang Early Childhood Teacher Training College. The results show that the VAE-GAN-DCR model performs well on three datasets, in which the Recall@20 value is increased by 12.15% and the NDCG@100 value is increased by 12.94% on the Movielens-1M dataset. The educational application experiment shows that the experimental group is significantly better than the control group in creative thinking activities and creative tendency, and the score of the total creative tendency scale reaches 2.61.The conclusions show that generative artificial intelligence technology can effectively improve the precision of educational resources recommendation and the development of students’ creativity, and provide a powerful support for the construction of smart classroom.
