Vadetis

Validator for Anomaly Detection in Time Series

Vadetis is web application to perform, compare and validate various anomaly detection algorithms using different configurations. It allows users to upload their own datasets as well as training data in order to perform outlier detection. The datasets can either be shared with other users or only be used by yourself. The datasets can be altered by injecting additional outliers. Look into the FAQ page for more information about the usage of this application.

Features

Main Parts:

Display
- Real-World Datasets
- Synthetic Datasets
Detection
- Real-World Datasets
- Synthetic Datasets
Recommendation
- Real-World Datasets
- Synthetic Datasets

Supported algorithms:

LISA (Pearson Correlation, Spatial Correlation, DTW)
Robust PCA (Huber Loss Function)
Histogram
Cluster (Gaussian Mixture)
Support Vector Machine
Isolation Forest

This application was implemented at the at the eXascale Infolab, a research group at the University of Fribourg, Switzerland.