PSO-LSSVM灰色组合模型在地下水埋深预测中的应用

针对LSSVM参数难以确定和单一方法预测精度不高的问题, 提出一种基于粒子群优化LSSVM灰色组合预测模型的学习方法. 利用粒子群算法的收敛速度快和全局优化能力, 优化LSSVM模型的惩罚因子和核函数参数. 避免了人为选择参数的盲目性. 在同一时刻利用不同长度序列的灰色预测方法对历史数据进行初步预测, 将初步预测结果的组合作为LSSVM的输入, 该时刻的实际值作为输出, 进行训练建立灰色LSSVM组合预测模型, 提高了模型的推广预测能力. 选取三江平原某地区1985年至2006年地下水埋深实测数据, 建立PSO-LSSVM组合预测模型. 通过两种方式对模型进行检验, 与其他模型相比, 该组合模型具有较高的预测精度.

LSSVM grey combined forecasting model based on PSO and its application in groundwater dynamic prediction

To solve the problems of the uncertain parameters of LSSVM and the low forecasting precision of single method, the learning algorithm of grey least squares support vector machines combined forecasting model optimized by particle swarm algorithm is proposed. Optimize two parameters of LSSVM model study by particle swarm algorithm's abilities of the fast convergence and whole optimization. It can escape from the blindness of man-made choice. First, the combinational results of initial forecasts are put as the input and the corresponding actual values are put as the output of LSSVM. Then we can get combinational model of the grey and the least squares support vector machine based on particle swarm algorithm by training it. The proposed combinational model can enhance the efficiency and the capability of forecasting. Actual data from 1985 to 2006 of area in Sanjiang plain is taken as the sample data. A combinational model based on PSO-LSSVM and GM(1,1) model is proposed. Predict precision of the model is examined by two ways, and the results show that it is more precise than the other methods.