SYSTEM AND METHOD FOR A DEVISED TRAINING PROCESS FOR IMBALANCED DATASET IN NON-INTRUSIVE ELEVATOR MONITORING

The present invention discloses a devised training method for imbalanced dataset in non-intrusive elevator monitoring, comprising the steps of: obtaining multi-variant signal data from non-intrusive current sensors (102); pre-processing the signal data (104); integrating the pre-processed signal data into a deep learning model (106); training the deep learning model by adopting algorithm to
balance the dataset (108); and monitoring condition and detecting anomaly of the elevator based on the deep learning model (110).