李鹏飞 王青青 毋建宏 樊怡彤

摘要 最小二乘支持向量机预测时，其参数的选取大部分只依赖于人工经验，无法实现自适应寻优，阻碍了其学习与泛化能力。针对该问题，采用灰狼优化算法对最小二乘支持向量机参数寻优，以1978—2016全国红枣产量数据为研究对象，利用最小二乘支持向量机的最优参数对红枣产量数据进行拟合与预测。为避免过拟合现象，将1978—2007和2013—2016年数据分别作为模型的训练与预测数据，2008-2012年数据用于交叉验证，同时为检验该模型的预测性能，将其与ARIMA模型的预测效果进行对比分析。实证分析表明，基于灰狼优化算法的最小二乘支持向量机模型预测的平均相对误差小于ARIMA模型预测的平均相对误差，其可适用于红枣产量的预测，也进一步表明灰狼优化算法对最小二乘支持向量机参数优化的有效性。

关键词 最小二乘支持向量机，全国红枣产量，灰狼优化算法，ARIMA

中图分类号 S126文献标识码 A文章编号 0517-6611（2020）06-0218-05

Abstract When predicting least squares support vector machine， most of its parameters are only dependent on artificial experience， and adaptive optimization cannot be achieved， which hinders its learning and generalization ability. To solve this problem， we used the grey wolf optimization algorithm to optimize the parameters of the least squares support vector machine， and took the 1978-2016 national jujube production data as the research object， and used the optimal parameters of the least squares support vector machine to calculate the red jujube yield data. In order to avoid overfitting， the data of 1978-2007 and 2013-2016 were used as the training and prediction data of the model， respectively. The data of 2008-2012 were used for crossvalidation. At the same time， it was combined with ARIMA to test the predictive performance of the model. The prediction effect of the model was compared and analyzed. The empirical analysis showed that the average relative error of the least squares support vector machine model based on the grey wolf optimization algorithm was smaller than the average relative error predicted by the ARIMA model， which could be applied to the prediction of jujube yield， and further indicated that the grey wolf optimization algorithm was effective to the least square support vector machine parameter optimization.

Key words Least squares support vector machine，National jujube yield，Grey wolf optimization algorithm，ARIMA

我国枣树资源十分丰富，在相当长的时间内，我国在世界红枣生产和贸易中占有绝对统治地位。红枣含有丰富的营养物质，在中医药学上有很高的实用价值。构建符合红枣产量变化的预测模型，科学准确预测红枣产量对巩固我国在世界红枣生產中的地位以及提升红枣产业经济效益具有重要的理论价值和实际意义。

随着预测理论的发展，BP神经网络[1]、支持向量机[2]、时间序列分析ARIMA[3]等传统以及相对应的改进模型被广泛应用于预测领域。……