Research on C5.0 Algorithm Improvement and the Test in Lightning Disaster Statistics
Shuonan Hou, Rongtao Hou, Xinming Shi, Jun Wang, Chengshang Yuan · International Journal of Control and Automation · 2014
After analysis the importance of continuous attributes in the C5.0 algorithm processing, and consider the shortcomings of discretization method of C5 algorithm, this paper proposes a new method based on Rough set theory-information entropy-discernible matrix discretization (RSIEDM).The method uses rough, information entropy and Discernible matrix that can be more reasonable and more accurately to continuously attribute discretization, and making created decision tree have better accuracy.In the application of optimization of lightning disaster statistics and evaluation result of lightning disaster, the algorithm which has obtained a better effect.