An in-depth study of how users sketch time series queries

Abstract

Data analysts and researchers actively work with time series data to gain insights about trends and patterns within a dataset. They employ various tools such as freehand sketches and digital software to identify these significant trends. However, this relies on users’ cognitive intelligence to accurately sketch the time series data they have in mind which in most cases is unreliable. Hence, the objective of this project is to study how users sketch time series data and their ability to accurately sketch the time series graph. A total of 48 participants have been engaged to sketch time series graphs of different characteristics. These include linear patterns, non-linear patterns, salient points, small fluctuations, and steep curves. Participants’ sketches are evaluated based on the sketch complexity and more importantly the similarity of their sketch to the actual time series displayed. This study found that participants can sketch linear straight-line graphs accurately and include salient points within their sketches. However, they are less accurate when it comes to time series graphs with multiple turning points or time series graphs that has a steep inconsistent rate of change between points. More research can be done with a larger group of participants to have a better overall view of the results. These results can be useful in developing Visual Query Systems that utilise users’ freehand sketches.