How to read these slides?

Navigate via arrow keys between slides Overview

Click on the menu bar items to navigate to chapters

Click here for PDF version
SCIENCE PASSION TECHNOLOGY

Guided Visual Analysis of Time Series Data
with Spiral Views and View Quality Measures

Stefanie Stoppacher, Julian Rakuschek,
Tobias Schreck

02.06.2025 - EuroVA

Given a long time series

Hard to spot repeating patterns through a line chart.

Alternative Visualization: Spirals

Weber et al. "Visualizing Time-Series on Spirals" (2001)

Not every number of segments per cycle is meaningful

12 segments per cycle
15 segments per cycle

How to find a good alignment? Guidance!

How does it work?

1. Algorithmically determine
cyclical phenomena

  • Discrete Fourier Transform (DFT)
  • Chi Square Periodogram (CSP)

2. Provide Visual Cues

Once aligned, we can compare cycles within a sector

Do segments within a sector change over time?

How do segments compare with neighboring segments?

Can we effectively guide users towards interesting sectors?

Selecting Sectors of the Spiral

Interactive Sector Selection

What makes a sector interesting?

  • Similarity between cycles
  • Monotonicity behavior
  • Change of values of over time

Quantifying these properties through quality measures

Monotonicity

  • Select a sector
  • For each cycle in the sector:
    • Count number of consecutively increasing elements
    • Normalize by cycle length
  • Average all cycle monotonicity scores to attain sector score

Dissimilarity

  • Select a sector
  • Compute pairwise Euclidean distance / DTW between all cycles (excluding self similarity)
  • Compute average distance

Visualization via heatmaps

We compute these quality measures for each sector configuration

The heatmaps are fully interactive

What are the benefits of this guidance approach?

Outliers may distort the color mapping

We recorded the sound level in our seminar room

Guidance unaffected

Euclidean Distance - Weekends rather quiet

Guidance unaffected

Euclidean Distance - During week rather busy

Guidance on large spirals

Daily temperature in Vienna over 20 years

Conclusion

Providing guidance on time series spirals may provide additional value

Our interface provides a framework for future extensions

Future Works

  • Additional quality measures
  • Multivariate time series
  • Searching for further use cases

Questions?

Slides:

Corresponding Author: Julian Rakuschek

julian.rakuschek@tugraz.at