Figure 10From: The shocklet transform: a decomposition method for the identification of local, mechanism-driven dynamics in sociotechnical time seriesJaccard similarity coefficients. Complimentary cumulative distribution function (CCDF) of Jaccard similarity coefficients for regions that Twitter’s ADV and our STAR algorithm detect patterns or anomalies (see Fig. 9). Window sizes are varied to include \(W_{s} \in \{0,3,5,7\}\) (i.e. detections within \(t_{i}\pm W_{s}\) are as part of the intersection). Time series with \(J_{\mathrm{word}_{i}}=0\) are omitted from the CCDF. The inset histogram shows the distribution of Jaccard similarity coefficients for \(W_{s}=0\) (i.e. exact matches), \(J=0\) time series are includedBack to article page