| 基于支持向量机回归的猪肺疫发病率预测模型研究 |
| |
| 引用本文： | 冯晓,乔淑,李国强,钱少俊,赵巧丽,周萌,胡峰,郑国清.基于支持向量机回归的猪肺疫发病率预测模型研究[J].河南农业科学,2016(1):138-142. |
| |
| 作者姓名： | 冯晓 乔淑 李国强 钱少俊 赵巧丽 周萌 胡峰 郑国清 |
| |
| 作者单位： | 1. 河南省农业科学院 农业经济与信息研究所,河南 郑州,450002;2. 河南省机关事务管理局,河南 郑州,450003 |
| |
| 基金项目： | 河南省重大科技专项，河南省农业科学院自主创新基金 |
| |
| 摘 要： | 为探明支持向量机回归(SVR)模型在动物疫病定量预测上的效果,以便为动物疫病防控决策提供依据,利用广西2007—2013年的猪肺疫月发病率时间序列,进行了SVR模型预测猪肺疫月发病率效果的研究。首先,以自相关函数法和Cao方法相结合,确定该时间序列的时间延迟为2,嵌入维数为6,并对其进行相空间重构;然后,依据主分量分析(PCA分布)方法判定该时间序列具有混沌特性,表明其在重构相空间中进行分析预测是可行的;最后,基于相空间重构结果构建SVR模型,分别采用网格搜索算法、遗传算法、粒子群算法对模型参数进行优化,并分析预测效果。结果表明,运用遗传算法优化SVR模型参数预测效果最优,平均绝对偏差(MAD)为0.043、均方误差(MSE)为0.003、平均绝对百分误差(MAPE)为0.202。可见,采用遗传算法优化的SVR模型对猪肺疫发病率的预测是可行有效的。
|
| 关 键 词： | 猪肺疫 预测 时间序列 支持向量机 相空间重构 |
Research on the Swine Lung Plague Incidence Forecast Model Based on Support Vector Machine Regression |
| |
| Abstract: | In order to explore the effect of support vector machine regression( SVR) model for quantita-tive forecast of animal disease and offer the decision basis for animal disease prevention and control,SVR model was used to forecast the monthly incidence of swine lung plague using the time series of monthly in-cidence of swine lung plague from 2007 to 2013 in Guangxi. Firstly,the auto correlation function method and Cao method was used to calculate the delay time as 2,the embedded dimension as 6,and the phase space of the time series was reconstructed. Secondly, principal component analysis method ( PCA ) was used to judge the chaotic characteristics of the time series,which showed the feasibility of the time series analyzed in the reconstructed phase space. Finally,based on the reconstructed phase space,SVR model was build to forecast the monthly incidence of swine lung plague,and the mesh optimization algorithm, particle swarm optimization algorithm and genetic algorithm was used to optimize the parameters of the model. The results showed that the optimal parameters of SVR model were optimized by genetic algorithm, and the mean absolute deviation(MAD) was 0. 043,the mean square error(MSE) was 0. 003,the mean absolute percent error(MAPE) was 0. 202. Therefore,GA-SVR model was feasible and effective for the forecast of swine lung plague incidence. |
| |
| Keywords: | swine lung plague forecast time series support vector machine phase space reconstruction |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
