paper

ASAP: Automatic Smoothing for Attention Prioritization in Streaming Time Series Visualization

  • Authors:
Read the paper

đź“ś Abstract

Time series visualization of streaming telemetry (i.e., charting of key metrics such as server load over time) is increasingly prevalent in recent application deployments. Existing systems simply plot the raw data streams as they arrive, potentially obscuring large-scale deviations due to local variance and noise. We propose an alternative: to better prioritize attention in time series exploration and monitoring visualizations, smooth the time series as much as possible to remove noise while still retaining large-scale structure. We develop a new technique for automatically smoothing streaming time series that adaptively optimizes this trade-off between noise reduction (i.e., variance) and outlier retention (i.e., kurtosis). We introduce metrics to quantitatively assess the quality of the choice of smoothing parameter and provide an efficient streaming analytics operator, ASAP, that optimizes these metrics by combining techniques from stream processing, user interface design, and signal processing via a novel autocorrelation-based pruning strategy and pixel-aware preaggregation. We demonstrate that ASAP is able to improve users’ accuracy in identifying significant deviations in time series by up to 38.4% while reducing response times by up to 44.3%. Moreover, ASAP delivers these results several orders of magnitude faster than alternative optimization strategies.

✨ Summary

Summary

The paper introduces ASAP, a streaming operator that automatically selects the smoothing window for time-series visualizations. Its objective is to remove short-term fluctuations while preserving large-scale deviations that should attract user attention.

ASAP uses a simple moving average and chooses the largest feasible smoothing window by optimizing two measures:

  • Roughness: the standard deviation of first differences, used as an inverse measure of visual smoothness.
  • Kurtosis preservation: a constraint requiring the smoothed series to retain at least the original series’ kurtosis, intended to prevent important extreme deviations from being averaged away.

The optimization problem is therefore to minimize the roughness of the smoothed series subject to a kurtosis-preservation constraint. For efficient execution, ASAP combines three techniques:

  1. Autocorrelation-based pruning: searches preferentially among window sizes associated with periodic structure and uses autocorrelation-derived roughness estimates to eliminate candidates.
  2. Pixel-aware preaggregation: uses the target display resolution to reduce the number of candidate windows and input points, making execution depend more on display resolution than on the original data volume.
  3. On-demand streaming updates: reuses prior search results and refreshes visualizations at configurable, human-observable intervals rather than after every incoming point.

In two user studies involving 270 participants, the authors report that ASAP-based visualizations improved anomaly-identification accuracy and reduced response time relative to raw, piecewise-aggregated, and oversmoothed alternatives. On the evaluated datasets, ASAP matched exhaustive search’s selected window while examining substantially fewer candidates. The combined optimizations produced improvements of several orders of magnitude, with the paper reporting approximately seven orders of magnitude over its baseline in one factor analysis.

Influence and subsequent use

The paper’s core criterion has been explicitly reused outside its original visualization setting. A later study on Bayesian spatial-model smoothing cites Rong and Bailis and adapts the roughness and kurtosis-preservation ideas to assess spatial under- and over-smoothing. (pmc.ncbi.nlm.nih.gov)

The Stanford DAWN project subsequently described ASAP as an automatic dashboard-smoothing technique and documented an ASAP.js implementation and demonstration, providing a direct software artifact associated with the work. (dawn.cs.stanford.edu) An open-source Rust implementation also identifies the paper as its algorithmic basis, indicating continued engineering reuse, although the available evidence does not establish broad commercial adoption. (ithub.global.ssl.fastly.net)