Visualet – a tool for visualizing shapelets, and exploring effective and interpretable ones

In the left upper part, the Load Dataset button is for loading a time series dataset and their class labels. The BSPCOVER button is for running our method on the loaded dataset. In the left lower part, the Load Shapelets button is for loading shapelets. After loading time series and shapelets, the time series chart panels can be used to visualize them, which can help demo attendees to have a quick overview of a dataset. Some basic information, such as the distance between time series and shapelets, and the class label of time series, are shown in the GUI. On top of the time series charts, the GUI presents some settings of Visualet, for instance, iteration number, and the alphabet size of SAX words, pCover number, the imputation methods for handling missing values (Impute) and the similarity metrics (ED). On the bottom of Visualet, the BSPCOVER Console will output log information.

The GUI screenshot of Visualet


We implemented the proposed demo in JAVA.

You can download Demo tool.

The Youtube link.


*G. Li, B. K. K. Choi, S. S. Bhowmick, K. Chun and G. L. Wong, S, Li, “Visualet: Visualizing Shapelets for Time Series Classification,” in CIKM2020, demo paper.

*G. Li, B. K. K. Choi, J. Xu, S. S. Bhowmick, K. Chun and G. L. Wong, “Efficient Shapelet Discovery for Time Series Classification,” in IEEE TKDE, doi: 10.1109/TKDE.2020.2995870.

HKBU Database Group