Detecting Anomalies in Sequential Data with Higher-order Networks
A major branch of anomaly detection methods relies on dynamic networks: raw sequence data is first converted to a series of networks, then critical change points are identified in the evolving network structure. However, existing approaches use first-order networks (FONs) to represent the underlying raw data, which may lose important higher-order sequence patterns, making higher-order anomalies undetectable in subsequent analysis. We present a novel higher-order anomaly detection method that is both parameter-free and scalable, building on an improved higher-order network (HON) construction algorithm. We show the proposed higher-order anomaly detection algorithm is effective in discovering variable orders of anomalies. Our data includes a synthetic 11 billion web clickstreams and a real-world taxi trajectory data.
PDF Abstract