摘要
BP网是神经网络时间序列预测方法中最常用的网络。针对BP算法局部搜索能力强,而遗传算法全局搜索优势突出的特点,将二者结合构造遗传BP神经网络,用于非平稳时间序列预测。仿真结果表明,该混合算法不仅提高了学习效率,而且对太阳黑子数预测的准确性高于BP算法、传统统计学预测方法。
BP network is commonly used in the forecasting of time series of neural networks. Combining the good local searching ability of BP algorithm and the strong global searching capacity of genetic algorithm, BP neural network of genetic algorithm is constructed and applied to forecast the nonstationary time series. Simulating tests show that the mixed algorithm can not only improve the learning efficiency, but also get higher precision than BP algorithm and traditional statistic methods in the forecasting of sunspot numbers.
出处
《四川理工学院学报(自然科学版)》
CAS
2008年第3期82-84,共3页
Journal of Sichuan University of Science & Engineering(Natural Science Edition)
关键词
BP神经网络
遗传算法
时间序列预测
BP neural network
genetic algorithm
forecasting of time series