支持向量机在短期用水量预测中的应用

针对基于经验风险最小化的神经网络存在模型结构较难确定和过学习的问题,根据时用水序列具有周期性和趋势性的特点,建立了基于支持向量机的时用水量预测模型.支持向量机采用结构风险最小化准则,在最小化学习误差的同时缩小模型泛化误差的上界,因此具有较强的泛化能力.此外,支持向量机通过将机器学习问题转化为二次规划问题,可获得全局最优解.实例分析结果表明,与基于BP网络的预测模型相比,基于支持向量机的时用水量预测模型建模速度更快,预测精度更高.

Application of support vector machine to predict short-term water consumption

Empirical Risk Minimization(ERM)-based neural network suffers drawbacks like overfitting the training data and the choice of the topology structure.According to the periodicity and trend of water demand series, an hourly water demand forecast model based on support vector machine(SVM) was developed.SVM possesses high generalization ability by employing structural risk minimization to minimize the learning errors and decrease the upper bound of prediction error.Furthermore,SVM converts machine learning problem into quadratic programming to achieve the global optimal solution.Case study showed that SVM-based hourly water demand prediction model performed significantly better than the BP neural network-based model on modeling speed and prediction.