Components aging, wear, and other faults will cause the elevator to run unsmoothly, resulting in vibration. However, conventional approaches rely on maintenance personnel to manually inspect the components and identify the root cause of the elevator vibration, which is inefficient. To improve the intelligence level of the inspection process, an automatic root cause location approach for elevator vibration based on Transformer and KNN is proposed. Firstly, the time series monitoring data is acquired by a three-axis acceleration sensor installed in the elevator cabin. Secondly, the collected data is pre-processed in the Transformer model. Specifically, the Transformer model encodes the timestamp information of the monitoring data into positional embedding and employs the multi-head self-attention mechanism for feature extraction. Then a vibration root cause classifier based on KNN is constructed, utilizing the extracted features as input for classification decision. Experimental results show that the accuracy of vibration root cause location can reach 94.57%, which proves the effectiveness of the proposed approach. Moreover, only one sensor installed in the elevator cabin is used, eliminating the need for additional sensors on other components, thereby enhancing the practical applicability of the approach.