| 摘要: |
| 【目的】 建立水文时间序列预测的核主成分支持向量机（KPCA_SVM）模型。【方法】 利用核主成分分析（KPCA）对输入数据进行非线性特征信息提取，并将提取的特征信息作为最小二乘支持向量机（LSSVM）的输入变量，建立KPCA_SVM预测模型。以甘肃民勤地区的月蒸发量为例，对模型的预测效果进行检验。【结果】 预测结果表明，KPCA_SVM模型预测效果优于PCA_SVM模型和LSSVM模型，预测平均相对误差为8.36%。【结论】 KPCA_SVM模型的预测效果优于没有特征提取的LSSVM模型。与主成分分析（PCA）提取特征相比，KPCA特征提取效果更好。 |
| 关键词: 水文时间序列 蒸发量 核主成分分析 支持向量机 KPCA_SVM模型 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金项目(50779052) |
|
| Establishment and application of hydrological time series forecasting model based on KPCA_SVM |
|
|
| Abstract: |
| 【Objective】 The KPCA_SVM model of hydrological time series forecasting model was established.【Method】 The method of Kernel Principle Component Analysis (KPCA) was used to obtain the feature information, and then the obtained series was used as input of Least Square Support Vector Machine model for forecasting. With monthly evaporation in the Minqin region as an example,it was applied to test forecasting result of model.【Result】 The results show that the KPCA_SVM model had a better effect on forecasting than PCA_SVM and LSSVM,and the average error was 8.36%.【Conclusion】 The forecasting effect of KPCA_SVM model was much better than that of LSSVM model without obtaining the feature information. In comparison with PCA,the performance of KPCA was better,too. |
| Key words: hydrological time series evaporation kernel principle component analysis support vector machine KPCA_SVM model |
