Unsupervised Offline Changepoint Detection Ensembles
Offline changepoint detection (CPD) algorithms are used for signal segmentation in an optimal way. Generally, these algorithms are based on the assumption that signalโs changed statistical properties are known, and the appropriate models (metrics, cost functions) for changepoint detection are used. Otherwise, the process of proper model selection can become laborious and time-consuming with uncertain results. Although an ensemble approach is well known for increasing the robustness of the individual algorithms and dealing with mentioned challenges, it is weakly formalized and much less highlighted for CPD problems than for outlier detection or classification problems. This paper proposes an unsupervised CPD ensemble (CPDE) procedure with the pseudocode of the particular proposed ensemble algorithms and the link to their Python realization. The approachโs novelty is in aggregating several cost functions before the changepoint search procedure running during the offline analysis. The numerical experiment showed that the proposed CPDE outperforms non-ensemble CPD procedures. Additionally, we focused on analyzing common CPD algorithms, scaling, and aggregation functions, comparing them during the numerical experiment. The results were obtained on the two anomaly benchmarks that contain industrial faults and failuresโTennessee Eastman Process (TEP) and Skoltech Anomaly Benchmark (SKAB). One of the possible applications of our research is the estimation of the failure time for fault identification and isolation problems of the technical diagnostics.
PDF AbstractCode
Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Benchmark |
---|---|---|---|---|---|---|
Change Point Detection | SKAB | Opt CPD algorithm (Mahalanobis metric) | NAB (standard) | 22.37 | # 4 | |
NAB (lowFP) | 19.9 | # 3 | ||||
NAB (LowFN) | 23.37 | # 4 | ||||
Change Point Detection | SKAB | OptEnsemble CPDE algorithm (WeightedSum+Rank) | NAB (standard) | 23.07 | # 3 | |
NAB (lowFP) | 20.52 | # 2 | ||||
NAB (LowFN) | 24.35 | # 3 | ||||
Change Point Detection | SKAB | BinSegEnsemble CPDE algorithm (WeightedSum+Rank) | NAB (standard) | 18.1 | # 7 | |
NAB (lowFP) | 15.36 | # 7 | ||||
NAB (LowFN) | 19.51 | # 6 | ||||
Change Point Detection | SKAB | WinEnsemble CPDE algorithm (Sum+MinAbs) | NAB (standard) | 19.38 | # 5 | |
NAB (lowFP) | 17.03 | # 5 | ||||
NAB (LowFN) | 20.35 | # 5 | ||||
Change Point Detection | SKAB | BinSeg CPD algorithm (Mahalanobis metric) | NAB (standard) | 24.1 | # 2 | |
NAB (lowFP) | 21.69 | # 1 | ||||
NAB (LowFN) | 25.04 | # 2 | ||||
Change Point Detection | SKAB | Win CPD algorithm (l1 metric) | NAB (standard) | 18.4 | # 6 | |
NAB (lowFP) | 16.22 | # 6 | ||||
NAB (LowFN) | 19.19 | # 7 | ||||
Change Point Detection | TEP | BinSegEnsemble CPDE algorithm (Min+MinMax/Rank) | NAB (standard) | 41.81 | # 1 | |
NAB (lowFP) | 41 | # 1 | ||||
NAB (LowFN) | 42.16 | # 1 | ||||
Change Point Detection | TEP | OptEnsemble CPDE algorithm (Min+MinMax/Rank) | NAB (standard) | 41.81 | # 1 | |
NAB (lowFP) | 41 | # 1 | ||||
NAB (LowFN) | 42.16 | # 1 | ||||
Change Point Detection | TEP | Opt CPD algorithm (Mahalanobis metric) | NAB (standard) | 36.88 | # 3 | |
NAB (lowFP) | 35.82 | # 3 | ||||
NAB (LowFN) | 37.29 | # 3 | ||||
Change Point Detection | TEP | Win CPD algorithm (Mahalanobis metric) | NAB (standard) | 27.79 | # 5 | |
NAB (lowFP) | 27 | # 5 | ||||
NAB (LowFN) | 28.05 | # 5 | ||||
Change Point Detection | TEP | BinSeg CPD algorithm (Mahalanobis metric) | NAB (standard) | 36.88 | # 3 | |
NAB (lowFP) | 35.82 | # 3 | ||||
NAB (LowFN) | 37.29 | # 3 | ||||
Change Point Detection | TEP | WinEnsemble CPDE algorithm (WeightedSum+MinAbs) | NAB (standard) | 25.14 | # 6 | |
NAB (lowFP) | 24.33 | # 6 | ||||
NAB (LowFN) | 26.29 | # 6 |