The Stipulator is a software tool for the processing of time series classification and prediction problems. It can be used for

  • the composition and analysis of sensor signal and time series data from various sources,
  • the formulation of time series classification and prediction requirements (aka "labelling of signals"),
  • the pre-processing and preparation of the time series data for the application of machine learning procedures.
  • Purpose
  • Features
  • Architecture

The collection and cleaning of the data is often one of the most tedious and extensive part of the work, when undertaking data mining studies. This is especially true, when the data comes from different sources of tests and experiments, when simulation data is mixed with test and/or artificial data, when signal are acquired with different measurement devices, etc. Unified and reproducible procedures are required then to assure conform and consistent data for comprehensive analysis and further processing. That's definitely what Stipulator is made for!

In addition the Stipulator provides routines for heuristic labelling of the data to define and manage the functional requirements for classification and prediction algorithms and models.

That way one can define

  • which action to take when and how
  • for which load cases
  • based on which sensor signals and signal preparation procedures

by the example based representation with tests and experimental data, simulation and artificial data.

This data then can be used for system tests and as base for the application of arbitrary machine learning procedures.


  • Definition and search of reference signal within the data base
  • Normalization of the signal data (e.g. by converting all the signal to a common unit)
  • Data plausibilization
  • Partially automated compensation and correction of faulty data
  • Definition and application of arbitrary sensor models
  • Definition and application of perturbation models for robustness assessments
  • Calculation of arbitrary characteristic numbers and attributes from the signal data
  • Definition and application of heuristics for the automated or manual definition of classification requirements (data labelling)
  • Execution of numerical conflict analysis to identify inconsistent classification requirements
  • Numerous functions for data analysis and visualization
  • Numerous distance metrics for signal comparisons
  • Conversion and export to other data formats
  • Extraction of the signal data including historic values to according matrices for the application of arbitrary machine learning procedures (e.g. in Brainer)

Technically, the Stipulator is a MATLAB Toolbox.

The mathematics of the Stipulator is built on the Signal Structure Toolbox from ANDATA, which comes as open software within the Stipulator distribution.

Optionally the Expectator can be integrated for the plausibilization of signals.

All functions of the Stipulator can also be triggered from an command line interface. That way all procedures can be composed to application specific scripts and apps to be integrated into customer specific environments.

MATLAB is registered trademark of The Mathworks, Inc.