智能组合预测方法及其应用

Intelligent Integration Forecasting Method and Its Application

由于具有能以任意精度逼近任意复杂非线性函数的优良性能,神经网络在灰色系统预测中得到了较大的应用。在已有的研究基础上,针对灰色神经网络进化时容易陷入局部最优,参数修正受阻的问题,建立基于遗传粒子群混合算法优化的新型灰色神经网络模型。首先将灰色神经网络进行数学建模,以便于优化算法的应用;其次,综合遗传算法与粒子群算法的优点,构造一种混合算法,运用混合算法对灰色神经网络进行优化;最后通过日本入华游客数量预测的算例研究,比较新型灰色神经网络与灰色神经网络、单一算法优化的灰色神经网络的预测精度。所得结果表明,混合算法优化的新灰色神经网络具有更好的预测性能,在社会经济领域有着广泛的应用前景。

Artificial neural network has been an important role in grey system prediction with the excellent properties having any arbitrary precision approximation for any nonlinear function. On the basis of existed research, considering problems of low efficiency, local optimum and retardation of parameter modification in grey neural network evolution process, in this paper a new grey neural network model is established based on genetic algorithm and particle swarm optimization. Firstly, a mathematical grey neural network is proposed in order to use optimization algorithm to solve it. Secondly, a hybrid algorithm is given to opti-mize the neural network model, which takes both advantages of genetic algorithm and particle swarm opti- mization. Finally, through calculation analysis of sample about tourist quantity forecasting Japan to China, the prediction accuracy of new grey neural network, grey neural network, genetic algorithm grey neural network and particle swarm optimization grey neural network is compared. The simulation results show that the new grey neural network based on genetic algorithm and particle swarm optimization has better forecast performance, which can hawa a wide application prospect in social and economic fields.