A class of problems between classification and regression, learning to predict ordinal classes, has not received much attention so far, even though there are many problems in the real world that fall into that category. Given ordered classes, one is not only interested in maximizing the classification accuracy, but also in minimizing the distances between the actual and the predicted classes. This paper provides a systematic study on the various methodologies that have tried to handle this problem and presents an experimental study of these methodologies with a cost sensitive technique that uses fixed and unequal misclassification costs between classes. It concludes that this technique can be a more robust solution to the problem because it minimizes the distances between the actual and the predicted classes, without harming but actually slightly improving the classification accuracy.
展开▼