Revisiting Interestingness Measures for Knowledge Discovery in Databases

Abstract

The voluminous amount of data stored in databases contains hidden knowledge which could be valuable to improve decision making process of any organization. As it is not humanely possible to analyze large databases, it has become essential to apply advanced data mining algorithms for extracting patterns (models) from data to support decision making. A number of data mining algorithms produce information of a statistical nature that allows the user to assess how accurate and reliable the discovered knowledge is? However, in many cases this is not enough for the users. Even if the discovered knowledge is highly accurate from a statistical point of view, it might not be interesting to the user. Therefore the process of knowledge discovery in databases (KDD) aims at discovering knowledge that is interesting and useful to the user. Most of the data mining algorithms so far have paid lot of attention to discovery of accurate and comprehensible knowledge. Though, the question of interestingness has been addressed time to time, it is being increasingly realized by data mining community that this subject needs a renewed focus. This paper is an attempt to review the measures of interestingness used in the data mining literature. The main contribution of the paper is to improve the understanding of interestingness measures for discovery of knowledge and identify the unresolved problems to set the directions for the future research in this area.