Data-driven construction of visual query interface for timeseries data

Abstract

This paper proposes an innovative way to define “hard-to-draw” patterns from time series data instances. These patterns are essentially time series shapelets, which are small, local patterns in a time series that are highly predictive of a class and very useful for visualization and summarization tasks. A GUI is then derived based on the pruning rules. It automatically generates hard-to-draw shapelets and display to user, which user can just simply drag to the canvas and complete the hand-drawn queries more efficiently.