基于遗传算法优选参数的灰色LS-SVM预测

Grey LS-SVM Forecasting with Parameter Optimized by Genetic Algorithm

摘要: 利用灰色预测方法中累加生成运算形成累加数据，将累加数据作为训练样本构造灰色LS-SVM，并利用遗传算法对灰色LS-SVM自身的参数进行优选，然后将基于遗传算法优选参数的灰色LS-SVM用于小样本预测。选取了典型例子进行验证，并与传统GM(1, 1)和LS-SVM方法进行对比。结果表明本文所提出的方法预测效果良好，且预测模型具有更好的泛化能力。

Abstract: This paper utilized the accumulation generation operation of grey prediction to produce accumulated data, and accumulated data were used to construct grey LS-SVM. At the same time the parameters for LS-SVM were pretreated through genetic algorithms to get the optimum parameter values, then the optimized LS-SVM based on genetic algorithms was used to small samples forecasting. A typical example was taken to be analyzed and compared with GM (1, 1) and LS-SVM method. The result shows that the method forecast effect is better, and the prediction model has better generalization ability.