Abstract: 【Purpose/significance】Compared with the retrospective detection of emerging themes, the prediction research on emerging themes can improve the accuracy and foresight of emerging theme identification, and help to enrich the method system of emerging theme detection and analysis.【Method/process】Firstly, the domain series topic is obtained by using the BERTopic model. Secondly, the model prediction feature set is constructed based on document frequency, citation frequency, Pscore and emerging scores. Then, based on the feature set data of the first three years of the topic, LSTM model is used to predict the emerging scores of the last two years, and the emerging topics in the field are judged. 【Result/conclusion】The emerging topic prediction method based on BERTopic and LSTM models was constructed, and an empirical study was conducted in the field of data security. The results showed that the emerging topics obtained by this method were more effective and reasonable through comparison with BP and SVM models and verifi⁃ cation of relevant data.【Innovation/limitation】Combining the characteristics of novelty, growth and impact, a single index emerging score is constructed to predict emerging themes, providing reference for future emerging theme prediction research, but without consid⁃ ering the prediction of possible future themes.