ADIA Lab and Crunch Lab Launch Structural Break Challenge

ADIA Lab, in partnership with Crunch Lab, successfully concluded its third machine learning competition: the ADIA Lab Structural Break Challenge.

The 2025 challenge invited participants to detect structural breaks in univariate time series data. Given a time series containing approximately 1,000 to 5,000 values and a designated point within it, participants were tasked with determining whether a structural break occurred at that point—indicating a change in the underlying data-generating process from that moment onwards.

Competitors had access to a large, labelled dataset of tens of thousands of time series for training. The problem carried significant relevance across domains such as climatology, finance, industrial monitoring, and healthcare, where detecting structural breaks can highlight critical events—from climate anomalies and equipment failures to market shifts and medical emergencies.

Participants explored a range of techniques, including statistical tests, feature extraction, time series modelling, and deep learning, to build robust detection algorithms.

The competition ran from 14 May to 15 September 2025, with a total prize pool of USD 100,000, including USD 40,000 for the overall winner and additional awards for the top ten entries.

The challenge attracted data scientists and machine learning researchers from around the world, contributing valuable insights toward advancing structural break detection in time series data and its real-world applications.