Cline: New multivariate decision tree construction heuristics

Abstract

Decision trees are often used in pattern recognition and regression problems. They are attractive due to high performance and easy-to-understand rules. Many different decision tree construction algorithms have been developed because of their popularity. In this work, we describe some new heuristic tree construction algorithms and test with 8 benchmark datasets. We compare the new method with other 21 tree induction algorithms. The results show that cline heuristics can be used in all types of classification problems because of its simplicity and acceptable performance.