Abstract

This document introduces further experimental results for the proposed dipm-SC method. First, we present our algorithm’s performance on synthetically generated time series datasets. We show the superior performance of the method over other possible shape-based extensions. Next, we present another real-world use-case scenario for GitHub repository dataset. We show by clustering multidimensional time-series of popularity of GitHub repositories coupled with other GitHub specific activities, we uncover underlying bursty and steady patterns of popularity gain similar to what we present in the original paper for Twitter hashtag dataset.

Date

October 19, 2025

Authors

Mert Ozer, Anna Sapienza, Andrés Abeliuk, Goran Muric, Emilio Ferrara