A comparative research on noise resistance for two heuristic algorithms

Decision tree induction is an important way of learning rules from examples. Due to the NP-hard problem, heuristic algorithms play a crucial role for generating short decision trees. This paper investigates the comparison between two heuristic algorithms in decision tree generation for the capacity of resisting noise. One heuristic is the well-known ID3 while the other is our previously proposed. The investigation is aiming at giving theoretically and experimentally some comparative advantages on the robustness for the two heuristics. Since most real world data are usually imprecise and inexact, the investigation to noise resistance is really necessary and significant to deal with the practical data in knowledge acquisition area.