| Citekey | YuEtAl2014Time |
| Source | own |
| Learning type | unsupervised |
| Input dimensionality | univariate |
- python 3
The algorithm uses windows around the current points to predict that point.
The difference between real and predicted value is used as anomaly score (k = window_size // 2).
The parameter k is the window size and therefore acts as a kind of smoothing factor.
The bigger the k, the smoother the predictions, the more values have big errors.
If k is too small, anomalies might not be found.
window_size should correlate with anomaly window sizes.
Small k
Big k
This parameter is only needed if the algorithm should decide itself whether a point is an anomaly.
It treats p as a confidence coefficient. It's the t-statistics confidence coefficient.
The smaller p is, the bigger is the confidence interval.
If p is too small, anomalies might not be found. If p is too big, too many points might be labeled anomalous.
Be aware that for example p = 0.05 == p = 0.95, as we are talking about percentiles of the t-statistics.
Small p
Big p
