Automatic decision support by rule exhaustion decision tree algorithm
| Autor: | Run-Zong Liu, Huiwu Luo, Bin Fang |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
Incremental decision tree
050208 finance Computer science business.industry Decision tree learning 05 social sciences Decision tree ID3 algorithm Classification Tree Method computer.software_genre Machine learning Classification rule 0502 economics and business Alternating decision tree Decision stump Artificial intelligence Data mining 050207 economics business computer |
| Zdroj: | 2016 International Conference on Wavelet Analysis and Pattern Recognition (ICWAPR). |
| DOI: | 10.1109/icwapr.2016.7731623 |
| Popis: | Decision tree is a landmark approach to automatic decision support. From a visualize decision tree, people can easily understand how the rules are produced. In this paper, the objective function is treated as an attribute of the classification rules. The proposed algorithm determines the value of the objective function firstly, and then determines other attribute values one by one to form a classification rule. A novel classification indicator is proposed which considers the samples deflection due to different attributes and the criterion of forming a classification rule. Based on this indicator, we could elect the superior attributes to classify the samples and discover new classification rules quickly and exhaustively. Synthetic random samples and rules are produced for experiment verification. The novel learning method shows its superiority to the classical ID3 decision tree in the experiments. The former not only spends less time and produces more rules than the latter, but also achieves a better precision than the latter. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|