I'm looking for ways to visualize large time-series other then in traditional 2-axis line charts. More specifically, I hope to find a way that intuitively reveals patterns in the data such as the spiral approach found in http://ieg.ifs.tuwien.ac.at/~aigner/teaching/ws06/infovis_ue/papers/spiralgraph_weber01visualizing.pdf
In my case, since I'm working with sensor data to track ones movements, days are a suitable interval. Another benefit of using spirals is that it doesn't require a linear increase in pixel space.
Other ideas: Overlaying data from multiple days on the same 24-hour x-axis or displaying the days individually, similar to a calendar.
Is there a better way? Any tips?
Best Answer
As you note, you could use an x axis showing 24 hours, then plot multiple days' data against this axis. This should reveal any intradaily patterns.
Alternatively, the x axis could run from Monday to Sunday, and you could plot each week's observations. This should show intraweekly (and intradaily) patterns; the weekend usually exhibits different patterns than the rest of the week.
Both are examples of seasonal plots; see here, also for other ideas for visualization.