基于遗传算法-支持向量机的水库叶绿素a浓度短期预测非线性时序模型

将支持向量机(SVM)法与遗传算法(GA)相结合，建立了基于GA-SVM的水库叶绿素a浓度非线性时间序列的短期预测模型。在建模过程中，采用遗传算法优化支持向量机的模型参数，同时利用相空间重构方法计算出时间序列的时间延迟和嵌入维数，确定出支持向量机的输入向量。最后将该模型用于对于桥水库的叶绿素a浓度时间序列进行短期预测。预测精度比单纯用人工神经网络方法有较大提高。

Genetic algorithm support vector machine model for short term prediction of chlorophyll a concentration nonlinear time series

The support vector machine (SVM) was combined with genetic algorithm (GA) to establish a model for predicting short term chlorophyll a concentration nonlinear time series. The genetic algorithm was used to optimize the parameters of the model. Meanwhile, the time delay and embedded dimension were calculated through phase reconstruction method. By this method the input vectors of the support vector machine can be defined. The model was applied to predict the chlorophyll a concentration in the Yuqiao Reservoir. The accuracy of the predicted result is much higher than that obtained by artificial neural network method.