Improved Ordinal Decision Tree Induction Algorithms

In the existing methods for inducing ordinal decision tree, ranking mutualinformation between the conditional attributes and the decision attribute is employedas a heuristic to select the extended attribute, while more and ordinal values ofattribute and correlation among the conditional attributes are not considered, this hasresulted in repeating an attribute for extended attributes. This will make classifierineffective and lower testing accuracy. In order to deal with this problem, the ordinaldecision tree induction algorithm is improved in this paper, the main works includethe following two aspects:1. A new ordinal decision tree induction algorithm is proposed. The extendedattribute selected with the proposed algorithm not only maximize the ranking mutualinformation between the candidate attributes and the decision attribute, but alsominimize the ranking mutual information between the candidate attributes and theselected conditional attributes on the same branch. Taking into account the correlationamong the conditional attributes can be avoided selecting the same one, and the ideasof the proposed method can really reflect the nature of the ranking mutualinformation. Compared with the existing algorithms, the proposed algorithm canimprove the test accuracy.2. In the ordinal data sets, some attribute values are more and ordinal; thoseattributes will affect the choice of extended attributes. For solving this problem,ranking mutual information rate instead of ranking mutual information to select theextended attributes. The proposed algorithm can overcome the problem mentionedabove and can improve the test accuracy.