Proximity-based density description with regularized reconstruction algorithm for anomaly detection

Jaehong Yu, Hyungrok Do

Abstract: Highlights•We propose a density-based OCC method, PDDRR, for anomaly detection.•We formulate the method as a density reconstruction subject to sparsity and smoothness.•The proposed method adeptly handles complex data distribution and robust to noise.•Experimental studies demonstrate our method outperforms the existing methods.