Abstract: features obscuring feature correlations，and the consequent impact on the area under the curve (AUC) in detection，this paper proposes an intelligent detection method based on Hermitian matrix decomposition of wireless sensor network (WSN) Internet of Things (IoT) data. WSN nodes are deployed at key elevator locations to collect operational data. Features are extracted via Hermitian matrix decomposition，followed by dimensionality reduction using kernel principal component analysis (KPCA). Finally，a random forest algorithm is adopted，which splits nodes based on the Gini coefficient and determines fault types through voting. Experimental results show that the proposed method regularizes the elevator operating frequency distribution and stabilizes amplitude fluctuations. Under different amounts of injected anomaly samples，the AUC values mostly remain above 0.90 with low fluctuation. Moreover， the method achieves more stable detection consistency and higher mean values under complex operating conditions. This approach is well-suited for intelligent detection of elevator emergency stop faults.