A mutual-information-based mining method for marine abnormal association rules

Long time series of remote sensing images are a key source of data for exploring large-scale marine abnormal association patterns, but pose significant challenges for traditional approaches to spatiotemporal analysis. This paper proposes a mutual-information-based quantitative association rule-mining algorithm (MIQarma) to address these challenges. MIQarma comprises three key steps. First, MIQarma calculates the asymmetrical mutual information between items with one scan of the database, and extracts pair-wise related items according to the user-specified information threshold. Second, a linking-pruning-generating recursive loop generates (k+1)-dimensional candidate association rules from k-dimensional rules on basis of the user-specified minimum support threshold, and this step is repeated until no more candidate association rules are generated. Finally, strong association rules are generated according to the user-specified minimum evaluation indicators. To demonstrate the feasibility and efficiency of MIQarma, we present two case studies: one considers performance analysis and the other identifies marine abnormal association relationships.