Detecting Data Errors with Employing Negative Association Rules

Ever increasing amount of data has led to the fact that the data quality has had a crucial impact on almost any IT applications. Nonetheless, data quality issues are nearly omnipresent. Recently, a great deal of researches has focused on improving data quality. Bad quality of the data can cause incorrect decision making. It is normally infeasible to guarantee sufficient data quality through manual inspection. Therefore (semi-)automatic data cleansing methods have to be employed. Data mining methods are ideal for this purpose, since they are aimed at finding abnormal patterns in large volumes of data. In this paper, we show the utility of negative association rules for detecting dubious data. The evaluations show that employing negative association rules provide interesting results for identifying incorrect data entries.