TimeEval Algorithms

Time Series Anomaly Detection Algorithms for TimeEval.

Description

This repository contains a collection of containerized (dockerized) time series anomaly detection methods that can easily be evaluated using TimeEval. Some of the algorithm's source code is access restricted and we just provide the TimeEval stubs and manifests. We are happy to share our TimeEval adaptations of excluded algorithms upon request, if the original authors approve this.

Each folder contains the implementation of an algorithm that will be build into a runnable Docker container using CI. The namespace prefix (repository) for the built Docker images is registry.gitlab.hpi.de/akita/i/.

Overview

Algorithm (folder)	Image	Language	Base image	Learning Type	Input Dimensionality
arima (restricted access)	registry.gitlab.hpi.de/akita/i/arima	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
autoencoder	registry.gitlab.hpi.de/akita/i/autoencoder	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
bagel	registry.gitlab.hpi.de/akita/i/bagel	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	univariate
baseline_increasing	registry.gitlab.hpi.de/akita/i/baseline_increasing	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
baseline_normal	registry.gitlab.hpi.de/akita/i/baseline_normal	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
baseline_random	registry.gitlab.hpi.de/akita/i/baseline_random	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
cblof	registry.gitlab.hpi.de/akita/i/cblof	python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
cof	registry.gitlab.hpi.de/akita/i/cof	python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
copod	registry.gitlab.hpi.de/akita/i/copod	python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
dae (DeNoising Autoencoder)	registry.gitlab.hpi.de/akita/i/dae	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
damp	registry.gitlab.hpi.de/akita/i/damp	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
dbstream	registry.gitlab.hpi.de/akita/i/dbstream	R 4.0.5	registry.gitlab.hpi.de/akita/i/r4-base	unsupervised	multivariate
deepant	registry.gitlab.hpi.de/akita/i/deepant	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
deepnap	registry.gitlab.hpi.de/akita/i/deepnap	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
donut	registry.gitlab.hpi.de/akita/i/donut	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	univariate
dspot	registry.gitlab.hpi.de/akita/i/dspot	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
dwt_mlead	registry.gitlab.hpi.de/akita/i/dwt_mlead	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
eif	registry.gitlab.hpi.de/akita/i/eif	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
encdec_ad	registry.gitlab.hpi.de/akita/i/encdec_ad	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
ensemble_gi	registry.gitlab.hpi.de/akita/i/ensemble_gi	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
fast_mcd	registry.gitlab.hpi.de/akita/i/fast_mcd	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
fft	registry.gitlab.hpi.de/akita/i/fft	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
generic_rf	registry.gitlab.hpi.de/akita/i/generic_rf	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	univariate
generic_xgb	registry.gitlab.hpi.de/akita/i/generic_xgb	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	univariate
grammarviz3	registry.gitlab.hpi.de/akita/i/grammarviz3	Java	registry.gitlab.hpi.de/akita/i/java-base	unsupervised	univariate
hbos	registry.gitlab.hpi.de/akita/i/hbos	python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
health_esn	registry.gitlab.hpi.de/akita/i/health_esn	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
hif	registry.gitlab.hpi.de/akita/i/hif	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	supervised	multivariate
hotsax	registry.gitlab.hpi.de/akita/i/hotsax	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
hybrid_knn	registry.gitlab.hpi.de/akita/i/hybrid_knn	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
if_lof	registry.gitlab.hpi.de/akita/i/if_lof	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
iforest	registry.gitlab.hpi.de/akita/i/iforest	python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
img_embedding_cae	registry.gitlab.hpi.de/akita/i/img_embedding_cae	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	univariate
kmeans	registry.gitlab.hpi.de/akita/i/kmeans	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
knn	registry.gitlab.hpi.de/akita/i/knn	python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
laser_dbn	registry.gitlab.hpi.de/akita/i/laser_dbn	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
left_stampi	registry.gitlab.hpi.de/akita/i/left_stampi	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
lof	registry.gitlab.hpi.de/akita/i/lof	python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
lstm_ad	registry.gitlab.hpi.de/akita/i/lstm_ad	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
lstm_vae	registry.gitlab.hpi.de/akita/i/lstm_vae	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	univariate
median_method	registry.gitlab.hpi.de/akita/i/median_method	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
mscred	registry.gitlab.hpi.de/akita/i/mscred	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
mstamp	registry.gitlab.hpi.de/akita/i/mstamp	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
mtad_gat	registry.gitlab.hpi.de/akita/i/mtad_gat	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
multi_hmm	registry.gitlab.hpi.de/akita/i/multi_hmm	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	supervised	multivariate
multi_subsequence_lof	registry.gitlab.hpi.de/akita/i/multi_subsequence_lof	python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
norma (restricted access)	registry.gitlab.hpi.de/akita/i/norma	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
normalizing_flows	registry.gitlab.hpi.de/akita/i/normalizing_flows	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	supervised	multivariate
novelty_svr	registry.gitlab.hpi.de/akita/i/novelty_svr	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
numenta_htm	registry.gitlab.hpi.de/akita/i/numenta_htm	Python 2.7	registry.gitlab.hpi.de/akita/i/python2-base	unsupervised	univariate
ocean_wnn	registry.gitlab.hpi.de/akita/i/ocean_wnn	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	univariate
omnianomaly	registry.gitlab.hpi.de/akita/i/omnianomaly	Python 3.6	registry.gitlab.hpi.de/akita/i/python36-base	semi-supervised	multivariate
pcc	registry.gitlab.hpi.de/akita/i/pcc	Python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
pci	registry.gitlab.hpi.de/akita/i/pci	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
phasespace_svm	registry.gitlab.hpi.de/akita/i/phasespace_svm	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
pst	registry.gitlab.hpi.de/akita/i/pst	R 3.5.2	registry.gitlab.hpi.de/akita/i/r-base
random_black_forest	registry.gitlab.hpi.de/akita/i/random_black_forest	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
robust_pca	registry.gitlab.hpi.de/akita/i/robust_pca	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
sand (restricted access)	registry.gitlab.hpi.de/akita/i/sand	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
sarima	registry.gitlab.hpi.de/akita/i/sarima	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
series2graph (restricted access)	registry.gitlab.hpi.de/akita/i/series2graph	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
s_h_esd	registry.gitlab.hpi.de/akita/i/s_h_esd	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
sr	registry.gitlab.hpi.de/akita/i/sr	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
sr_cnn	registry.gitlab.hpi.de/akita/i/sr_cnn	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch	semi-supervised	univariate
ssa (restricted access)	registry.gitlab.hpi.de/akita/i/ssa	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	univariate
stamp	registry.gitlab.hpi.de/akita/i/stamp	R 3.5.2	registry.gitlab.hpi.de/akita/i/tsmp -> registry.gitlab.hpi.de/akita/i/r-base	unsupervised	univariate
stomp	registry.gitlab.hpi.de/akita/i/stomp	R 3.5.2	registry.gitlab.hpi.de/akita/i/tsmp -> registry.gitlab.hpi.de/akita/i/r-base	unsupervised	univariate
subsequence_fast_mcd	registry.gitlab.hpi.de/akita/i/subsequence_fast_mcd	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	univariate
subsequence_knn	registry.gitlab.hpi.de/akita/i/subsequence_knn	python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
subsequence_if	registry.gitlab.hpi.de/akita/i/subsequence_if	python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
subsequence_lof	registry.gitlab.hpi.de/akita/i/subsequence_lof	python 3.7	registry.gitlab.hpi.de/akita/i/pyod -> registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
tanogan	registry.gitlab.hpi.de/akita/i/tanogan	python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch -> registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
tarzan	registry.gitlab.hpi.de/akita/i/tarzan	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-torch	semi-supervised	univariate
telemanom	registry.gitlab.hpi.de/akita/i/telemanom	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	semi-supervised	multivariate
torsk	registry.gitlab.hpi.de/akita/i/torsk	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	multivariate
triple_es	registry.gitlab.hpi.de/akita/i/triple_es	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
ts_bitmap	registry.gitlab.hpi.de/akita/i/ts_bitmap	python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	univariate
valmod	registry.gitlab.hpi.de/akita/i/valmod	R 3.5.2	registry.gitlab.hpi.de/akita/i/tsmp -> registry.gitlab.hpi.de/akita/i/r-base	unsupervised	univariate

Usage

Prerequisites

You need the following tools installed on your development machine:

• git
• docker
• access to this repository

Please make yourself familiar with the contents of this repository and read this document carefully!

Testing an algorithm and its TimeEval integration

Testing an algorithm locally can be done in two different ways:

1. Test the algorithm's code directly (using the tools provided by the programming language)
2. Test the algorithm within its docker container

The first option is specific to the programming language, so we won't cover it here.

Each algorithm in this repository will be bundled in a self-contained Docker image so that it can be executed with a single command and no additional dependencies must be installed. This allows you to test the algorithm without installing its dependencies on your machine. The only requirement is a (x86-)Docker runtime. Follow the below steps to test your algorithm using Docker:

1. Prepare base image You'll need the required base Docker image to build your algorithm's image. If you find yourself situated in the HPI network (either VPN or physically), you are able to pull the docker images from our docker repository registry.gitlab.hpi.de/akita/i/. If this is not the case you have to build the base images yourself as follows:

   • change to the 0-base-images folder: cd 0-base-images
   • build your desired base image, e.g. docker build -t registry.gitlab.hpi.de/akita/i/python3-base:0.2.5 ./python3-base
   • (optionally: build derived base image, e.g. docker build -t registry.gitlab.hpi.de/akita/i/pyod:0.2.5 ./pyod)
   • now you can build your algorithm image from the base image (see next item)

2. Build algorithm image Next, you'll need to build the algorithm's Docker image. It is based on the previously built base image and contains the algorithm-specific source code.

   • Change to the root directory of the timeeval-algorithms-repository.
   • build the algorithm image, e.g. docker build -t registry.gitlab.hpi.de/akita/i/lof ./lof

3. Train your algorithm (optional) If your algorithm is supervised or semi-supervised, execute the following command to perform the training step:

   mkdir -p 2-results
   docker run --rm \
       -v $(pwd)/1-data:/data:ro \
       -v $(pwd)/2-results:/results:rw \
   #    -e LOCAL_UID=<current user id> \
   #    -e LOCAL_GID=<current groupid> \
     registry.gitlab.hpi.de/akita/i/<your_algorithm>:latest execute-algorithm '{ \
       "executionType": "train", \
       "dataInput": "/data/dataset.csv", \
       "dataOutput": "/results/anomaly_scores.ts", \
       "modelInput": "/results/model.pkl", \
       "modelOutput": "/results/model.pkl", \
       "customParameters": {} \
     }'

   Be warned that the result and model files will be written to the 2-results-directory as the root-user if you do no pass the optional environment variables LOCAL_UID and LOCAL_GID to the container.

4. Execute your algorithm Run the following command to perform the execution step of your algorithm:

   mkdir -p 2-results
   docker run --rm \
       -v $(pwd)/1-data:/data:ro \
       -v $(pwd)/2-results:/results:rw \
   #    -e LOCAL_UID=<current user id> \
   #    -e LOCAL_GID=<current groupid> \
     registry.gitlab.hpi.de/akita/i/<your_algorithm>:latest execute-algorithm '{ \
       "executionType": "execute", \
       "dataInput": "/data/dataset.csv", \
       "dataOutput": "/results/anomaly_scores.ts", \
       "modelInput": "/results/model.pkl", \
       "modelOutput": "/results/model.pkl", \
       "customParameters": {} \
     }'

   Be warned that the result and model files will be written to the 2-results-directory as the root-user if you do no pass the optional environment variables LOCAL_UID and LOCAL_GID to the container.

Additional information

TimeEval base Docker images

To benefit from Docker layer caching and to reduce code duplication (DRY!), we decided to put common functionality in so-called base images. The following is taken care of by base images:

• Provide system (OS and common OS tools)
• Provide language runtime (e.g. python3, java8)
• Provide common libraries / algorithm dependencies
• Define volumes for IO
• Define Docker entrypoint script (performs initial container setup before the algorithm is executed)

Please consider the folder 0-base-images for all available base images.

TimeEval algorithm interface

TimeEval uses a common interface to execute all the algorithms in this repository. This means that the algorithms' input, output, and parametrization is equal for all TimeEval algorithms.

Execution and parametrization

All algorithms are executed by creating a Docker container using their Docker image and running it. The base images take care of the container startup and they call the main algorithm file with a single positional parameter. This parameter contains a String-representation of the algorithm configuration JSON. Example parameter JSON (2021-02-03):

{
  "executionType": 'train' | 'execute',
  "dataInput": string,   # example: "path/to/dataset.csv",
  "dataOutput": string,  # example: "path/to/results.csv",
  "modelInput": string,  # example: "/path/to/model.pkl",
  "modelOutput": string, # example: "/path/to/model.pkl",
  "customParameters": dict
}

Custom algorithm parameters

All algorithm hyper parameters described in the paper are exposed via the customParameters configuration option. This allows us to set those parameters from TimeEval.

Attention!

TimeEval does not parse a manifest.json file to get the custom parameters' types and default values. We expect the users of TimeEval to be familiar with the algorithms, so that they can specify the required parameters manually. However, we require each algorithm to be executable without specifying any custom parameters (especially for testing purposes). Therefore, please provide sensible default parameters for all custom parameters within the method's code.

Please add a manifest.json file to your algorithm anyway to aid the integration into UltraMine and for user information.

If your algorithm does not use the default parameters automatically and expects them to be provided, your algorithm will fail during runtime if no parameters are provided by the TimeEval user.

Input and output

Input and output for an algorithm is handled via bind-mounting files and folders into the Docker container.

All input data, such as the training dataset and the test dataset, are mounted read-only to the /data-folder of the container. The configuration options dataInput and modelInput reflect this with the correct path to the dataset (e.g. { "dataInput": "/data/dataset.test.csv" }).

All output of your algorithm should be written to the /results-folder. This is also reflected in the configuration options with the correct paths for dataOutput and modelOutput (e.g. { "dataOutput": "/results/anomaly_scores.csv" }). The /results-folder is also bind-mounted to the algorithm container - but writable -, so that TimeEval can access the results after your algorithm finished. An algorithm can also use this folder to write persistent log and debug information.

Temporary files and data of an algorithm are written to the current working directory (currently this is /app) or the temporary directory /tmp within the Docker container.

Example calls

The following Docker command represents the way how TimeEval executes your algorithm image:

docker run --rm \
    -v $(pwd)/1-data:/data:ro \
    -v $(pwd)/2-results:/results:rw \
#    -e LOCAL_UID=<current user id> \
#    -e LOCAL_GID=<groupid of akita group> \
#    <resource restrictions> \
  registry.gitlab.hpi.de/akita/i/<your_algorithm>:latest execute-algorithm '{ \
    "executionType": "train", \
    "dataInput": "/data/dataset.csv", \
    "modelInput": "/data/model.pkl", \
    "dataOutput": "/results/anomaly_scores.ts", \
    "modelOutput": "/results/model.pkl", \
    "customParameters": {} \
  }'

This is translated to the following call within the container:

mkdir results
docker run --rm \
    -v $(pwd)/1-data:/data:ro \
    -v $(pwd)/2-results:/results:rw \
  registry.gitlab.hpi.de/akita/i/<your_algorithm>:latest bash
# now, within the container
<python | java -jar | Rscript> $ALGORITHM_MAIN '{ \
  "executionType": "train", \
  "dataInput": "/data/dataset.csv", \
  "modelInput": "/data/model.pkl", \
  "dataOutput": "/results/anomaly_scores.ts", \
  "modelOutput": "/results/model.pkl", \
  "customParameters": {} \
}'