Sustained Evolving Co-occurrence Pattern Mining

Discovering spatio-temporal co-occurrence patterns is an important task in numerous domains. Existing studies on the co-occurrence pattern mining solely focus on identifying the sustained emerging patterns while neglecting the sustained disappearing patterns. Thus, the evolution of co-occurrence patterns over time cannot be completely captured. To solve this problem, we introduce the concept of sustained evolving co-occurrence patterns (SECP) to consider both sustained emerging and disappearing patterns. To measure the evolution of SECP over time, we propose the evolving participation index incorporating the impact of the emergence of new instances and the disappearance of existing instances on the pattern evolution. Moreover, we propose an efficient SECP mining algorithm based on the changed participation-instance search, which can solve the problem of high time overhead of traditional co-occurrence pattern mining algorithms. On both real and synthetic spatio-temporal datasets, extensive experiments demonstrate the effectiveness and the efficiency of the proposed method.