The main objective of this repository is to organize the results obtained by the Anomaly and Outlier Detection (AOD) method that works on time series data.
The Instances directory stores the instance files in csv that were generated and used in the experiments. The published files are about multivariate time series. The instances of YOB (Yahoo Outlier Benchmark) and NAB (Numenta Anomaly Benchmark) are not disclosed here, but can be obtained from the links YOB and NAB. The instance of the case study was not disclosed due to industrial property issues.
The Results directory stores the results obtained by the AOD method. This directory has the following subdirectories: 1-YOB, 2-NAB, 3-Synthetic-Dataset and 4-Case-Study. See the directory structure shown below.
The 1-YOB directory has four subdirectories which are: A1, A2, A3 and A4. Inside each of these directories we have the images showing the results of the experiments on the YOB benchmark. The images give a photographic idea that shows how the method works. Some of the images from the experiments are shown below.
Results - YOB | Results - YOB |
---|---|
A1 Benchmark - Instance 55 | A2 Benchmark - Instance 38 |
A3 Benchmark - Instance 39 | A4 Benchmark - Instance 35 |
The 2-NAB directory has seven subdirectories which are: artificialNoAnomaly, artificialWithAnomaly, realAdExchange, realAWSCloudwatch, realKnownCause, realTraffic and realTweets. Inside each one of these directories we have the images showing the results of the experiments on the NAB benchmark. Some of the images from the experiments are shown below.
The 3-Synthetic-Dataset directory has five files in which these files contain the images showing the results of experiments on synthetic data generated using the agots tool. Some of the images from the experiments are shown below.
The 4-Case-Study directory has a file containing an image showing the result of one of the experiments performed on the case study with multivariate data. An example image of one of the experiments is shown below.