v3.0 [+EMeriTAte with Python bindings]

This version is associated with the following paper:

G. Bergami, Emma Packer, Kirsty Scott, Silvia Del Din. "Predicting Dyskinetic Events through Verified Multivariate Time Series Classification". IDEAS 2024

Differently from version v2.9, where the pipeline was scattered throughout different invocations of Python and C++ code, this version wraps the minimum constituents of KnoBAB for EMeriTAte via pybind11, while preserving the original C++ codebase in C++.

After cloning this repository, you can easily install the Python bindings for the EMeriTAte classifier using pip3 as follows:

pip3 install EMeriTAte/

The present codebase provides an example for using the wrapper, as well as the legacy version of the pipeline.

Legacy pipeline

The legacy pipeline considered originally for the paper included the following steps:

1. Raw data pre-processing in Python, plus DT mining -- Python

2. Specification mining and SAT checking (also phase 4) -- C++

3. Specification crawling for preparing the SAT phase across logs -- Python

4. SAT checking (see Phase N°2)

5. Competitor testing over the same dataset