Research On Ordinal Classification And Regression Of Interval Valued Data

Traditional decision tree algorithms for interval valued data only can deal with non-ordinal classification problems. In order to solve this problem, an algorithm is presented to solve the ordinal classification problems, where bot h the condition attributes with interval values and the decision attribute meet the monotonic requirement. The algorithm uses the rank mutual information to select extended attributes, whic h guarantees that the output decision tree is monotonic. An example is put forward to illustrate the induction process of a monotonic decision tree.On the basis of the above work, another algorithm is proposed in this paper to solve the ordinal regression problems, where the condition attributes are ordinal interval values and the decision attribute is continuous. The heuristic adopted in this algorithm combines the variance and rank mutual information, which can guarantee both the dispersion degree of the decision attribute value and the monotonic consistency degree between condition attributes and decision attribute are considered.