| Sand-Dust Storm Ensemble Forecast Model Based on Rough Set |
| |
| 作者姓名： | 路志英 杨乐 李艳英 赵智超 |
| |
| 作者单位： | School of Electrical Engineering and Automation Tianjin University，School of Electrical Engineering and Automation，Tianjin University，School of Electrical Engineering and Automation，Tianjin University，School of Electrical Engineering and Automation，Tianjin University，Tianjin 300072，China，Tianjin 300072，China，Tianjin 300072，China，Tianjin 300072，China |
| |
| 摘 要： | To improve the accuracy of sand-dust storm forecast system, a sand-dust storm ensemble forecast model based on rough set (RS) is proposed. The feature data are extracted from the historical data sets using the self-organization map (SOM) clustering network and single fields forecast to form the feature values with low dimensions. Then, the unwanted attributes are reduced according to RS to discretize the continuous feature values. Lastly, the minimum decision rules are constructed according to the remainder attributes, namely sand-dust storm ensemble forecast model based on RS is constructed. Results comparison between the proposed model and the back propagation neural network model show that the sand-storm forecast model based on RS has better stability, faster running speed, and its forecasting accuracy ratio is increased from 17.1% to 86.21%.
|
| 关 键 词： | 粗糙集 同频率区间 判定表 特征抽出 |
Sand-Dust Storm Ensemble Forecast Model Based on Rough Set |
| |
| LU Zhiying,YANG Le,LI Yanying,ZHAO Zhichao.Sand-Dust Storm Ensemble Forecast Model Based on Rough Set[J].Transactions of Tianjin University,2007,13(6):441-446. |
| |
| Authors: | LU Zhiying YANG Le LI Yanying ZHAO Zhichao |
| |
| Institution: | School of Electrical Engineering and Automation, Tianjin University, Tianjin 300072, China |
| |
| Abstract: | To improve the accuracy of sand-dust storm forecast system, a sand-dust storm ensemble forecast model based on rough set (RS) is proposed. The feature data are extracted from the historical data sets using the self-organization map (SOM) clustering network and single fields forecast to form the feature values with low dimensions. Then, the unwanted attributes are reduced according to RS to discretize the continuous feature values. Lastly, the minimum decision rules are constructed according to the remainder attributes, namely sand-dust storm ensemble forecast model based on RS is constructed. Results comparison between the proposed model and the back propagation neural network model show that the sand-storm forecast model based on RS has better stability, faster running speed, and its forecasting accuracy ratio is increased from 17.1% to 86.21%. |
| |
| Keywords: | rough set equal frequency intervals decision table feature extraction |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
