Change point detection in time series data with random forests

• Published in: Control engineering practice Vol. 18; no. 8; pp. 990 - 1002
• Main Authors: Auret, Lidia, Aldrich, Chris
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01-08-2010; Elsevier
• Subjects: Applied sciences; Artificial intelligence; Computer science; control theory; systems; Detection algorithms; Dynamical systems; Dynamics; Equivalence; Exact sciences and technology; Forests; Learning and adaptive systems; Machine learning; Mathematical models; Monitoring; Singular value decomposition; Subspace methods; Time series; Time series analysis; Trajectories; Time series analysis; Detection algorithms; Subspace methods; Machine learning; Singular value decomposition; Event detection; Complex system; Random series; Non linear control; Binary data; Embedding; Change point; Dynamical system; Modeling; Aggregate model; Time analysis; Phase space; Pattern extraction; Monitoring; Binary classification; Time series; Subspace method; Pattern recognition; Non linear system; Operating conditions; Surveillance; Feature extraction; Artificial intelligence
• ISSN: 0967-0661; 1873-6939
• DOI: 10.1016/j.conengprac.2010.04.005

A large class of monitoring problems can be cast as the detection of a change in the parameters of a static or dynamic system, based on the effects of these changes on one or more observed variables. In this paper, the use of random forest models to detect change points in dynamic systems is considered. The approach is based on the embedding of multivariate time series data associated with normal process conditions, followed by the extraction of features from the resulting lagged trajectory matrix. The features are extracted by recasting the data into a binary classification problem, which can be solved with a random forest model. A proximity matrix can be calculated from the model and from this matrix features can be extracted that represent the trajectory of the system in phase space. The results of the study suggest that the random forest approach may afford distinct advantages over a previously proposed linear equivalent, particularly when complex nonlinear systems need to be monitored.