最小二乘支持向量机用于水量预测

Water Consumption Prediction Using Least Squares Support Vector Machines Based on Immune Clone

针对标准支持向量机建模时间长的缺点,为了城市用水量准确预测,需建立有效的预测模型。采用的最小二乘支持向量机基于结构风险最小化,并在支持向量机的基础上,将求解二次规划问题转化线性方程组,采用径向基核函数,使最小二乘支持向量机模型的待定参数比标准支持向量机少,可大大加快建模速度,同时还采用了人工免疫系统的自适应动态克隆选择算法,在寻优过程中能够准确、快速地搜索最小二乘支持向量机的最优参数。把上述模型用于城市日用水量预测,具有学习速度快,也具有良好的非线性建模和泛化能力,而且预测精度较高。

Due to the time consumption of the standard support vector machine modeling, an effective prediction model needs to be established. The least squares support vector machines （LSSVM） can deal with the relation between empirical risk minimization principle and machine learning VC dimensions based on the principle of structure risk minimization. Moreover , on the basis of support vector machine （ SVM ）, LSSVM translate the reprogramming question into linear equations, and radial kernels are adopted, thus making the model of LSSVM have less parameters than standard SVM, greatly accelerate the speed of modeling. Owing to introducing adaptive dynamic clone selection algorithm （ADCSA） based on artificial immunity in this paper, the optimal parameters of least squares support vector machines can accurately and quickly found in the algorithm. The model is applied to the prediction of daily water consumption in the city. The simulation results show that this prediction model has the characteristics of rapid training, non - linear modeling, generalization and higher accuracy.