Query-constraint-based association rule mining from diverse clinical datasets in the national sleep research resource

Abstract

Secondary use of biomedical data has gained much attention recently to facilitate rapid knowledge discovery in biomedicine. Association Rule Mining (ARM) has been a popular technique for biomedical researchers to perform exploratory data analysis and discover potential relationships among variables in biomedical datasets. However, ARM of a high-dimensional biomedical dataset may produce a large number of rules that may not be interesting. In this paper, we introduce a query-constraint-based ARM (QARM) approach for exploratory analysis of diverse clinical datasets integrated in the National Sleep Research Resource (NSRR), which enables the rule mining on a subset of data containing items of interest based on a query constraint. In addition, biomedical datasets always contain semantically similar variables, thus we performed similar-variable-merging so that rules with simlar variables are not obtained. Applying QARM on five datasets from NSRR obtained a total of 6,921 rules with a minimum confidence of 60% (using top 50 rules for each query constraint).