Abstract: To improve the recognition accuracy of traditional intelligent lost circulation recognition models due to limited lost circulation samples, the long short-term memory (LSTM) network and auto-encoder (AE) are combined to produce an integrated LSTM-AE based lost circulation recognition model. Firstly, multiple LSTM-AE models with different numbers of hidden neurons are trained using normal samples, and several base recognizers with better recognition performance were selected using the reconstruction scores. Subsequently, the recognition results of these base recognizers were fused using ensemble learning. This approach mitigates the tendency that a single model is prone to producing false positives and false negatives due to overlearning local sample characteristics, thus enhances the overall recognition accuracy. The integrated LSTM-AE model was trained and tested using 6000 sets of stand pipe pressure, outlet flow and mud pit volume data from 18 wells under normal drilling conditions in an oil field. The results show that the proposed method achieves a recognition accuracy of 94.7%, which is superior to the recognition results of other commonly used intelligent models and provides a new way for lost circulation monitoring.