| 杨 静，张大斌，方洁凤，李培汉.基于优化 EEMD 和支持向量机的国内大豆价格预测[J].广东农业科学,2019,46(11):134-140 |
|
查看全文
HTML
基于优化 EEMD 和支持向量机的国内大豆价格预测 |
| Domestic Soybean Price Forecast Based onEEMD and Support Vector Regression |
| |
| DOI：10.16768/j.issn.1004-874X.2019.11.019 |
| 中文关键词: EEMD 平行延拓 支持向量回归 大豆价格预测 |
| 英文关键词: EEMD parallel extension Support Vector Regression soybean price forecast |
| 基金项目:广东省自然科学基金（2016A030313402）；广东省大学生创新创业项目（201810564104） |
|
| 摘要点击次数: 2261 |
| 全文下载次数: 720 |
| 中文摘要: |
| 【目的】国内大豆的价格易受到多种因素的影响，具有非线性等特点，很难进行准确的预测。为
了提高预测精度，提出一种优化的 EEMD-SVR 集成预测方法。【方法】为解决 EMD 分解中存在的模态混叠和
端点效应问题，使用 EEMD 和平行延拓法结合的优化方法，加入白噪声并在原始序列两端延拓出多个极值，将
大豆原始价格分解为多个 IMF 分量，从而使数据趋于平稳。运用支持向量回归（SVR）算法对各分量进行预测，
引入遗传算法寻找参数最优解，对各分量的预测结果进行再次集成，重构大豆市场价格预测值。【结果】为了
检验优化组合模型的预测效果，采取多种模型进行比较，结果发现预测指标 MSE、RMSE、MAPE 都有明显提高。
【结论】采用优化的 EEMD 分解算法和支持向量机的组合模型，可以有效抑制 EMD 分解的端点效应和模态混叠
问题，相对于其他传统预测模型，预测效果更好。 |
| 英文摘要: |
| 【Objective】The price of domestic soybean is easily influenced by many factors, which is characterized
by non-linearity, and it is difficult to make accurate prediction. In order to improve the prediction accuracy, an optimized
EEMD-SVR integrated prediction method is proposed.【Method】For solving the problems of modal aliasing and endpoint
effect in EMD decomposition, by using the optimization method of EEMD and parallel extension method, white noise was
added and multiple extreme values were extended at both ends of the original sequence, and the original soybean price was
decomposed into multiple IMF components. In this way, the data tended to be stabilized. The Support Vector Regression
(SVR) algorithm was used to predict each component, the genetic algorithm was introduced to find the optimal solution of
parameters, the prediction results of each component were re-integrated, and the market price prediction value of soybean
was reconstructed.【Result】In order to test the prediction effect of the optimized combination model, a variety of models
were compared, and the results showed that the prediction indicators MSE, RMSE and MAPE were improved significantly.
【Conclusion】The combined model of the optimized EEMD decomposition algorithm and support vector machine can
effectively suppress the endpoint effect and modal aliasing of EMD decomposition. And the prediction effect is better than
that of other traditional prediction models. |
|
查看/发表评论 下载PDF阅读器 |
