New index and algorithm find strengthening co-occurrence patterns in changing spatial data

Liuwei Li, Peizhong Yang, Lizhen Wang and Hongmei Chen have proposed a method for mining sustained emerging spatio-temporal co-occurrence patterns (SECOPs) in dynamic spatial databases, published in Data Science and Engineering. These are sets of features whose instances increasingly appear close together in space and time.

The standard participation index can miss such patterns when new instances change neighbour relationships or when features grow at different rates. The authors introduce a weighted emerging participation index (WEPI), which uses a Gaussian-kernel weighting of emerging intensity. Because WEPI is not antimonotone, they prove an upper bound that allows candidate pruning, and build IM-DFS, an incremental depth-first mining algorithm.

Tests used point-of-interest data from Beijing and Shanghai for 2014 to 2021. IM-DFS ran faster than two adapted baseline algorithms, and without pruning the search exceeded 24 hours. Users must still set distance and threshold values.

Why it matters: Analysts working with continuously growing point datasets, such as health events, facilities or ecological records, get a way to detect strengthening co-location that classical measures miss, with lower computational cost. The study itself is not Africa-specific.