A Study on the Evolution and Hotspot Prediction of Research Topics in Artificial Intelligence Using BERTopic and LSTM Approaches

With the rapid development of artificial intelligence technology and the continuous deepening of interdisciplinary integration, accurately identifying and predicting research hotspots in this field is of great value for grasping the direction of disciplinary development and optimizing the allocation of scientific research resources. Existing research methods have obvious limitations in terms of topic identification accuracy and evolutionary prediction capabilities, making it difficult to effectively process massive amounts of literature data and capture complex temporal features. This paper combines BERTopic topic modeling with LSTM time series prediction methods. First, BERTopic is used to identify research topics, followed by the construction of a prediction model using LSTM to forecast the evolution and hotspots of topics in the field of artificial intelligence. The research findings not only provide new methodological tools for predicting disciplinary development but also offer important data support and theoretical foundations for scientific research management decisions.