Earthquake Time Series Prediction Based on GA- Locally Weighted Support Vector Regression

摘要：

There exists obvious shortages in earthquake prediction area by using traditional methods. In this paper, combined with the phase space reconstruction, locally weighted support vector regression and genetic algorithm, we proposed a new support vector regression model based on genetic algorithm and locally weighted regression algorithm. Using the new method we successfully performed prediction on a chaotic time series of earthquake. Experiments show that through changing SVM risk function by the introduction of locally weighted regression and the introduction of genetic algorithm to optimize the key parameters of SVM, Support vector machine with chaotic time series data on the earthquake has good generalization ability. The algorithm has better effect than the single use of locally weighted regression or support vector regression.

关键词： phase space reconstruction (PSR) locally weighted regression (LWR) support vector regression (SVR) genetic algorithm (GA) earthquake time series prediction

作者: LUO Sheng ZHANG Ping-wei WU Shao-chun XIE Huang

作者单位: School of Computer Engineering and Science Shanghai University ShangHai, China

会议类型: 国际会议

会议名称: 2010 International Conference on Circuit and Signal Processing(2010年电路与信号处理国际会议 ICCSP 2010)

会议地点: 上海

会议语种:英文

页码: 253-256

在线出版日期: 2010-12-25（万方平台首次上网日期，不代表论文的发表时间）