UChicago Alum John Paparrizos Honored with SIGMOD Test-of-Time Award for Advancing Time Series Analytics
Every financial tick and hospital monitor produces a flickering line of data—signals that, taken together, reveal the invisible patterns shaping daily life. Thanks to the work of UChicago alumn John Paparrizos, making sense of these time-based patterns, or data points recorded at specific time intervals, has become far more accurate and efficient. This year, Paparrizos received the SIGMOD Test-of-Time Award for a method called k-Shape, first published during his Ph.D. studies at Columbia University. Paparrizos continued his research in the area during his postdoctoral fellowship at the University of Chicago under the guidance of Distinguished Professor Michael Franklin and Associate Professor Aaron Elmore. This method helps researchers and practitioners in fields ranging from healthcare to finance identify connections in their data that were once easy to overlook. The honor by SIGMOD recognizes research done 10-12 years prior that shows a profound impact over the intervening decade.
Every minute, hospitals generate streams of heart rate data, and financial markets churn out charts mapping the rise and fall of stocks. Behind the scenes, Professor John Paparrizos’s work makes sense of these patterns—helping doctors diagnose and investors forecast with greater accuracy. In his paper, k-Shape: Efficient and Accurate Clustering of Time Series, Paparrizos introduces a new and improved method, called k-Shape, that automatically clusters temporal data, such as daily stock market prices or patient heart rates. What makes k-Shape unique is that data is clustered based on its overall shape or pattern, rather than just its quantitative value. The k-Shape algorithm is highly efficient at identifying data that exhibits similar behavior, such as stocks that rise and fall in tandem. Previous methods employed distance metrics, such as Euclidean distances, which are highly sensitive to distortions in time-series data, including amplitude and phase differences. Paparrizos and his co-author, Luis Gravano, show that the k-Shape method is more accurate and efficient than older techniques and can be used in many different fields.
While it was already a seminal paper at the time, for Paparrizos, receiving the Test of Time award for his very first PhD paper carries special meaning.
“At the time, the ideas were not always aligned with prevailing trends, but I believed deeply in their importance,” He reflected. “Receiving this award a decade later is both a personal milestone and a reminder that foundational research can have enduring impact.”
As an Assistant Professor, Paparrizos’ research now lies at the intersection of data management and machine learning, with a focus on enabling effective analysis of unstructured data such as time-series and sensor-generated data from physical devices, equipment, and sensors. He is particularly excited about working across the full analytics stack, from low-level components like compression and indexing, all the way through the high-level machine learning inference tasks such as anomaly detection, clustering, forecasting, and classification.
“My interest in this area developed organically during my early research years,” Paparrizos recalled. “I was drawn to the growing disconnect between the richness of real-world data and the limitations of existing systems to analyze it effectively and scalably. I found that time series and unstructured data were especially underserved, despite being central to many domains. That led me to pursue research that combines rigor with practical relevance, aimed at building methods and systems that are robust, efficient, and widely applicable.”
Building on this foundation, Paparrizos and his research group continue to push the boundaries of analytics over unstructured and time series data. Their recent efforts are focused on democratizing access to complex data, building tools that make information more accessible and analyzable, particularly in domains where ground truth data is limited or costly. A major focus remains on developing robust, adaptive methodologies that can respond to changes in data and operate under uncertainty, a feature that is critical for unsupervised learning tasks that remain underexplored. His group is also engaged in research on vector databases, which are increasingly important across a wide range of applications, including modern foundation models. One of those key directions is designing new compression and indexing methods that do not compromise accuracy while achieving substantial speedups. Paparrizos hopes that these contributions would support scalable and reliable inference in real-world systems.
Paparrizos is excited about the future applications of his research, and he is actively pursuing collaborations with industry partners to ground his research in real-world use cases.
“I’m particularly interested in applications in manufacturing, infrastructure monitoring, and healthcare analytics—areas where timely and accurate insights from complex data can make a significant impact,” he said. “By aligning our research with real challenges faced by practitioners, we aim to ensure that our contributions translate into meaningful, deployable solutions. These collaborations also provide valuable feedback loops that guide our ongoing work.”
Looking ahead, Paparrizos is working on building foundation models for time series analytics. He is excited to pursue these topics with emphasis on both methodological innovation and system-level optimization.
To learn more about Paparrizos’ work, visit his website here.
