Granger causality-based cluster sequence mining for spatio-temporal causal relation mining

Abstract We proposed a method to extract causal relations of spatial clusters from multi-dimensional event sequence data, also known as spatio-temporal point process. The Granger cluster mining algorithm identifies the pairs data that have causality over time with each other. It extended algorithm, which utilized statistical inference technique identify occurrence relation, based on causality. In addition, utilizes false discovery rate procedure control significance Based experiments both synthetic and semi-real we confirmed is able multiple different sets even when disturbed high level noise. False helps increase accuracy more under such case make less-sensitive hyperparameters.