Abstract: Rolling bearings are one of the key components in metro elevator systems. To improve the diagnostic efficiency for potential faults of elevator rolling bearings, this paper proposes a K-Nearest Neighbors algorithm based on Locality-Sensitive Hashing. By using LSH to bucket similar data, data dimensionality reduction is achieved to enhance classification efficiency; the reduced-dimension data is then classified by the KNN network. Simultaneously, a data acquisition system for fault signals of metro elevator rolling bearings was developed, and the data collected by this equipment was used to validate the effectiveness of the classification algorithm. Experiments demonstrate that the proposed method can effectively identify different types of rolling bearing faults and shows significant advantages compared to the traditional KNN.