COPPER - Constraint optimized prefixspan for epidemiological research

Abstract

Sequential pattern mining, is a data mining technique used to study the temporal evolution of events describing a complex phe- nomenon. This technique has a limited application due to the high number of common sequences generated by dense datasets. To tackle this problem, we propose COP, an extension of the PrefixSpan algorithm oriented towards optimizing the relevance of the results obtained in the sequential patterns mining process. Indeed, we use multiple and simultaneous constraints that represent the expertise of researchers in a specific domain. Experiments conducted on datasets associated to dengue epidemic monitoring show an improve in result relevance from an expert's point of view, as well as, a considerable speed gains for mining dense datasets.