Motivation

Throughout our time in college we were faced with many different types of time series data. As visual analysis is a very big part of the initial process of understanding the behavior of the data at hand, we always plotted the data, and did so using code. This was always quite a time-consuming process, especially when we wanted to view our data in more than one way. We were surprised at the lack of online sources that allowed users to easily explore time series data, so we decided to create our own tool that does just that.

Goal

Our goal is to create an interactive tool that will allow users to visualize and explore different types of time series data regardless of its structure.
The data could be uni- or multi-variate, could include tens of points or tens of thousands, and these points could form all types of trends, patterns, or correlation.

Implementation

To make our tool as inclusive as possible, we first turned to academia to get inspiration for the types of visualizations we should start with. The following two papers caught our eye:

The two main advantages of these two visualizations are:

  1. Although they are both designed for large datasets, a little bit of tweaking can make them very useful for small ones as well. A dense lines chart can easily become a normal line graph, while a horizon graph can turn into a small multiples one.

  2. They guide viewers towards different behaviors of their data. Dense lines are good for understanding where data overlaps, while horizon charts make it easy understand how different time series relate to each other.
The entire site has been written in Javascript with Bootstrap and D3.

Future Steps

We aim to add more interactivity to the current graphs, provide information about the behavior of the uploaded data (mean, variance, stationarity, etc), improve the loading speed, and add more types of visualization.
If you have any suggestions on things you would like to see in this page, we would love to hear them! Just contact us on the links below.

Contact

Petra Kumi: pkumi@wpi.edu
Philippe Lessard: plessard@wpi.edu