A Java implementation of twitter anomaly detection. But optimized the time complexity from o(n^2) to o(nlogn).

[1. Vallis, O., Hochenbaum, J. and Kejariwal, A., (2014) “A Novel Technique for Long-Term Anomaly Detection in the Cloud”,
 6th USENIX Workshop on Hot Topics in Cloud Computing, Philadelphia, PA.]
(https://www.usenix.org/system/files/conference/hotcloud14/hotcloud14-vallis.pdf)
[2. Rosner, B., (May 1983), “Percentage Points for a Generalized ESD Many-Outlier Procedure”, Technometrics, 25(2), pp. 165-172.]
[3. STL: A Seasonal-Trend Decomposition Procedure Based on Loess](http://www.wessa.net/download/stl.pdf)

• STL-java reference (https://github.com/brandtg/stl-java), but we implements the stl in java as stlplus described, faster and can handle NA values (some data used stl-java will throw some exception).
• Twitter-anomalyDetection (https://github.com/twitter/AnomalyDetection), we optimize the algorithm from o(n^2) to o(nlogn)
• STL Test and Anomaly Detection Test is from (https://anomaly.io/anomaly-detection-twitter-r/):
  • use R stl decompose as stl test benchmark.
  • use R twitter anomaly detection as anomaly detection test benchmark
  • this lib performs as well as twitter. And it can even find the anomalies those could not be detected by Twitter-anomalyDetection!

For more information please read code, the ReadMe will update later!