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 4.2.0	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
grammarviz3_multi	registry.gitlab.hpi.de/akita/i/grammarviz3_multi	Java	registry.gitlab.hpi.de/akita/i/java-base	unsupervised	multivariate
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_norma (restricted access)	registry.gitlab.hpi.de/akita/i/multi_norma	Python 3.7	registry.gitlab.hpi.de/akita/i/python3-base	unsupervised	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
mvalmod	registry.gitlab.hpi.de/akita/i/mvalmod	R 4.2.0	registry.gitlab.hpi.de/akita/i/tsmp -> registry.gitlab.hpi.de/akita/i/r4-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 4.2.0	registry.gitlab.hpi.de/akita/i/r4-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 4.2.0	registry.gitlab.hpi.de/akita/i/tsmp -> registry.gitlab.hpi.de/akita/i/r4-base	unsupervised	univariate
stomp	registry.gitlab.hpi.de/akita/i/stomp	R 4.2.0	registry.gitlab.hpi.de/akita/i/tsmp -> registry.gitlab.hpi.de/akita/i/r4-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 4.2.0	registry.gitlab.hpi.de/akita/i/tsmp -> registry.gitlab.hpi.de/akita/i/r4-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 concepts of TimeEval, and read the TimeEval documentation and 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 (examples assume that you want to build the image for the LOF algorithm):

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:

   1.1 change to the 0-base-images folder:

   cd 0-base-images

   1.2 build your desired base image:

   docker build -t registry.gitlab.hpi.de/akita/i/python3-base:0.2.5 ./python3-base

   1.3 (conditional step) because lof depends on a derived base image, we need to additionally build the pyod-image

   • re-tag the base image python3-base with latest (because the pyod-image depends on the latest python3-base-image):

   docker tag registry.gitlab.hpi.de/akita/i/python3-base:0.2.5 registry.gitlab.hpi.de/akita/i/python3-base:latest

   • build intermediate image:

   cd ../1-intermediate-images
   docker build -t registry.gitlab.hpi.de/akita/i/pyod:0.2.5 ./pyod

   1.4 now you can build your algorithm image from the base image (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

   cd ..
   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 (not necessary for LOF):

   mkdir -p results
   docker run --rm \
       -v $(pwd)/data:/data:ro \
       -v $(pwd)/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 results-directory as the root-user if you do not 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 results
   TIMEEVAL_ALGORITHM=lof
   docker run --rm \
       -v $(pwd)/data:/data:ro \
       -v $(pwd)/results:/results:rw \
   #    -e LOCAL_UID=<current user id> \
   #    -e LOCAL_GID=<current groupid> \
     registry.gitlab.hpi.de/akita/i/${TIMEEVAL_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 results-directory as the root-user if you do not pass the optional environment variables LOCAL_UID and LOCAL_GID to the container.