A1311
Title: Online change point detection in high-dimensional vector auto-regressive models
Authors: Abolfazl Safikhani - George Mason University (United States) [presenting]
Abstract: Online change point detection consists of sequentially monitoring a time series and raising the alarm if a shift in the data distribution is detected. Given data generated by a high-dimensional vector auto-regressive model, an algorithm is proposed to detect changes in the model transition matrices in an online format. The algorithm consists of two main steps. First, the estimation of transition matrices and variance of error terms are calculated by applying regularization methods to the training data. As new batches of data are observed, a specific test statistic is calculated in the second step to check whether the transition matrices have changed. Asymptotic normality of the test statistic in the regime of no change points is established under mild conditions. The alarm will be raised if the test statistic is larger than a certain quantile of standard normal distribution. Further, the relationship between the power of the test and jump size is established, and it is verified that for a large enough jump size, the power of the test converges to one. The proposed algorithm is memory-saving since it only requires storing the estimations and new batches of data in computer memoralgorithm's effectivenessgorithm is confirmed empirically through various simulation settings and comparisons with some competing methods. Finally, applications to analyze shocks in S&P 500 data and to detect the timing of seizures in EEG data are discussed.
