How Can Clusters of Population Trajectories be Identified? Comparing the Potential of Dynamic Time Warping and Sequence Analysis

ABSTRACT Reducing complexity while retaining the maximum amount of information is a key challenge for population analysis. Solving this challenge becomes a necessity when looking at numerous areas over extended periods, which defy manual pattern recognition efforts. This paper introduces Dynamic Time Warping (DTW) as a novel method for population time series clustering, capable of creating distinct, well‐separated groups for process‐centered population analysis. DTW is benchmarked against a Sequence Analysis model and established typologies based on size, location or density with population data from nearly 3000 towns in Germany for the period 2001 to 2022. The results indicate that DTW consistently outperforms the Sequence Analysis model across various cluster quality measures, producing better‐separated typologies of population trajectories. Both models are highly superior to the established typologies. The results highlight the significant advantages of using DTW for clustering continuous time series data, making it well‐suited for identifying typologies of municipal population trends.