Identifying insightful salinity and temperature variations in ocean data
Yo-Ping Huang, Li-Jen Kao, Frode Eika Sandnes · 2008
Global ocean salinity and temperature variations are attracting increasing attention as scientists are trying to extract more knowledge from collected ocean data to better understand global change. Association rules mining can be applied to ocean salinity and temperature data to discover spatial-temporal patterns that reveal salinity and temperature variations. Since we are addressing the associations of salinity/temperature events among different time and locations, the events can be grouped into clusters before mining starts, and the discovered association rule that has its antecedent and consequent from different clusters will be of most interest. However, are the discovered rules important or insightful? In this paper, an importance measurement for association rules with antecedent and consequent from different clusters is proposed. The importance measurement quantifies the rule’s antecedent and consequent impact to their clusters, respectively. A rule is insightful if its importance measurement is above a predefined threshold. The insightful rules can then be presented to experts for further analysis.