Association Rules to Predict the Likelihood of Recoverability of an Attribute in a Database

摘要

Missing data values are a common problem in certain databases. In our paper, we propose a method that uses association rules, a data mining technique, to predict the likelihood of recoverability of missing values of a particular attribute in a database. Our method has several important applications. For example, using our proposed method, vulnerable attributes in terms of recovering their missing values can be identified. Hence, those attributes can be paid more attention in data collection process to reduce me amount of missing data values in those attributes. In addition, application-type data recovery methods can be developed for those particular attributes to recover their missing values.