DiscreeTest

Description

DiscreeTest, is a two-step evaluation metric for ranking discretization methods for time-series data. The underlying assumption of DiscreeTest is that an optimal discretization method should preserve the dynamic patterns observed in the original data across all variables.

Availability

The source code of DiscreeTest is available in our Github repository: https://github.com/VeraLiconaResearchGroup/Benchmarking_TSDiscretizations

Preprint

Y Li, T Jann, P Vera-Licona. (2018). Benchmarking Time-Series Data Discretization on Inference Methods. Submitted.

Related