Using weighted hybrid discretization method to analyze climate changes

Abstract

Data mining is the process of posing queries to large quantities of data and extracting information, often previously unknown, using mathematical, statistical and machine learning techniques. However some of the data mining techniques like classification and clustering cannot deal with numeric attributes though most real dataset contains some numeric attributes. Continuous attributes should be divided into a small distinct range of nominal attributes in order to apply data mining techniques. Correct discretization makes the dataset succinct and contributes to the high performance of classification algorithms. Meanwhile, several methods are presented and applied, but it is often dependent on the area. In this paper, we propose a weighted hybrid discretization technique based on entropy and contingency coefficient. Also we analyze performance evaluation with well-known techniques of discretization such as Equal-width binning, 1R, MDLP and ChiMerge. © 2012 Springer-Verlag.