Time-series shapelets are discriminative subsequences, recently found effective for time series classification (TSC). It is evident that the quality of shapelets is crucial to the accuracy of TSC. However, the majority of research has focused on building accurate models from some shapelets candidates. To determine such shapelets candidates, some existing studies are surprisingly simple. For example, there are studies that enumerate subsequences of some fixed lengths, or randomly select some subsequences as shapelets candidates. The major bulk of computation is then on building the model from the candidates, which is computationally costly. Hence, we propose a novel efficient shapelets discovery method, called BSPCOVER, to discover a set of high-quality shapelets candidates for model building.

The overview of BSPCOVER



We implemented the proposed algorithms in JAVA. A well-known benchmark of time series classification datasets, UCR Time Series Classification Archive, has been tested.

Source code can be found in here.
The demo page is here

The style of password is “xxx-xxxxxxxx”,
all the characters are with lower case,
two parts are linked by “-“,
the first part is first letter of last three words in the website head title,
the second part is the name of our method.

Accuracy of UCR Archive datasets

All accuracy of UCR Archive of BSPCOVER is here.

Every day is a new day. :)



Thanks the research community for supporting the datasets.

HKBU Database Group