Greedy online change point detection
This addresses the issue of outlier sensitivity in change point detection for time series analysis, though it appears incremental as it builds on existing CPD frameworks with a new optimization approach.
The paper tackles the problem of high false discovery rates in online change point detection methods by proposing Greedy Online Change Point Detection (GOCPD), which finds change points by maximizing data probability from concatenated independent models. The method achieves logarithmic complexity via ternary search for single change points and demonstrates effectiveness on synthetic and real-world data.
Standard online change point detection (CPD) methods tend to have large false discovery rates as their detections are sensitive to outliers. To overcome this drawback, we propose Greedy Online Change Point Detection (GOCPD), a computationally appealing method which finds change points by maximizing the probability of the data coming from the (temporal) concatenation of two independent models. We show that, for time series with a single change point, this objective is unimodal and thus CPD can be accelerated via ternary search with logarithmic complexity. We demonstrate the effectiveness of GOCPD on synthetic data and validate our findings on real-world univariate and multivariate settings.