An Effective Tree-Based Algorithm for Ordinal Regression

Recently ordinal regression has attracted much interest in machine learning. The goal of ordinal regression is to assign each instance a rank, which should be as close as possible to its true rank. We propose an effective tree-based algorithm, called Ranking Tree, for ordinal regression. The main advantage of Ranking Tree is that it can group samples with closer ranks together in the process of tree learning. This approach is compared with original decision tree. Experiments on some synthetic and real-world datasets show that Ranking Tree outperforms original decision tree in terms of speed and accuracy as well as robustness.