Invited Talk

Temporal Pattern Discovery for Scalable and Interpretable Time Series Analysis

Assistant Professor Len Feremans, Hasselt University, Belgium

BIO

I am an Assistant Professor at Hasselt University and a member of the Data Science Institute. My research focuses on data mining and machine learning, with an emphasis on developing fair, explainable, and practically useful AI methods. My main research interests include pattern and motif mining, time series analysis—such as anomaly detection and classification—and recommender systems.

Abstract

Time series motif discovery and sequential pattern mining study similar questions from different perspectives: how can recurring temporal structure be found efficiently and represented meaningfully?

In this talk, I will connect both fields through several lines of work. Earlier research used mined sequential patterns for interpretable anomaly detection and time series classification.

More recent work, including MotiPlus and MotiSet, focuses on discovering high-quality and non-redundant motif sets within time series, with ongoing extensions toward motifs of varying length.

I will also discuss FRM-Miner, which applies frequent pattern mining to motif discovery across large collections of time series.

Together, these results point toward a broader view of temporal pattern discovery in which motifs and sequential patterns provide a common foundation for scalable, interpretable time series analysis.