Comparative Study of Association Rule Mining and MiSTIC in Extracting Spatio-temporal Disease Occurrences Patterns

Extracting interesting and useful patterns from spatio-temporal datasets is more difficult than extracting the corresponding patterns from traditional numeric and categorical data due to the complexity of spatial data types and the embedded topologies, spatial and temporal relationships, and spatial autocorrelation. The objective of epidemiology is to identify disease causes and correlating them to spatially explicit disease patterns and variations in health risks. The main issue in traditional mining of association rules in disease surveillance data is that a large number of rules are discovered, but most of them are of limited use in addressing the stated objectives or original questions asked. Moreover, not all of the generated rules are interesting (due to their inability to conclusively mine spatio-temporal prevalence and causative factors of diseases), and some rules may be ignored. These drawbacks result as these methods ignore the inherent spatio-temporal dependency in such data. This paper makes a case for the use of MiSTIC algorithm to address these issues, compare the use of traditional association rule mining in context of Salmonellosis disease management, and share new insights. An illustrative case study presented here suggests that in comparison to traditional association rule mining, even simple spatio-temporal data mining approaches taking into consideration the spatio-temporal interdependencies in disease data, can provide new and valuable scientific insights towards efficient disease surveillance and management.