Research on Dynamic Spatial Evolution Model of Traditional Villages Based on Deep Learning

In the process of modernization, the spatial form of traditional villages is undergoing significant changes, and their cultural values and ecosystems are facing severe challenges. Based on deep learning technology, this paper explores the dynamic spatial evolution of traditional villages, combines convolutional neural network (CNN) and long short-term memory network (LSTM), constructs a time series prediction model, and uses generative adversarial network (GAN) to simulate the spatial pattern of villages under future policy scenarios. The results show that the building density increases year by year, while the green space coverage decreases significantly, and the two are negatively correlated. The CNN-LSTM model has high prediction accuracy, and the scenario simulation generated by GAN provides scientific basis for policy making. This paper provides technical support and optimization path for the protection and sustainable development of traditional villages.