Analyzing and Mining Ordered Information Tables

Work in inductive learning has mostly been concentrated onclassifying. However, there are many applications in which itis desirable to order rather than to classify instances. For modellingordering problems, we generalize the notion of information tablesto ordered information tables by adding order relations inattribute values. Then we propose a data analysis model byanalyzing the dependency of attributes to describe the propertiesof ordered information tables. The problem of mining orderingrules is formulated as finding association between orderings ofattribute values and the overall ordering of objects. An orderingrules may state that “if the value of an object x on anattribute a is ordered ahead of the value of another object yon the same attribute, then x is ordered ahead of y”. Formining ordering rules, we first transform an ordered informationtable into a binary information table, and then apply any standardmachine learning and data mining algorithms. As an illustration,we analyze in detail Maclean's universities ranking for the year2000.