Spatial Data Science for Regional Pattern Analysis: Dynamic Time Warping-Based Clustering of East Java’s Economic Indicators

Motivated by the need to better capture dynamic regional disparities, this study examines spatial and temporal development patterns in East Java, Indonesia, using spatial panel data from 2020 to 2023. A data-driven framework is proposed that integrates Principal Component Analysis (PCA) for dimensionality reduction, Dynamic Time Warping (DTW) for temporal similarity measurement, and spatially constrained clustering using the SKATER algorithm. PCA compresses multiple socio-economic indicators, GDP growth, GDP level, Human Development Index (HDI), and population density, into a unified development profile, enabling comparison of regional trajectories over time. DTW captures non-linear temporal alignment, while SKATER preserves spatial coherence in cluster formation. The resulting clusters are used to construct an endogenous spatial weight matrix that reflects functional regional relationships rather than purely geographic adjacency. Validation using Moran’s I indicates stronger spatial autocorrelation compared to conventional contiguity-based weights, suggesting improved representation of spatial interaction. Four clusters reveal distinct development patterns and uneven regional trajectories. By integrating dimensionality reduction with temporal alignment and spatial clustering, the proposed approach extends dynamic spatial weighting toward a functional interpretation of regional dependence and offers a transferable framework for spatial data science and regional policy analysis.