摘要
为对未来电信业务总量和各类用户数进行有效预测,通过分析历年电信业务总量和各类用户数,建立小波神经网络预测模型,以提高预测精度。在神经网络预测模型建立中,神经网络中的转移函数使用小波函数替代,从而得到小波基神经网络系统;通过对自适应学习速度和参数初始值选取的改进,获得高几率初始参数并加快算法收敛速度。
It is necessary to give more efficient method for forecasting the estimate telecommunication.A model based on Wavelet Neural Network is introduced in telecommunications to forecast the income and the uesers.To build wavelet basis NN(Neural Network),the sigmoid function is replaced with the wavelet in NN,and adaptive learning rate and initialization of parameters are also used to obtain high probability and high convergence speed.When a model of wavelet neural network is established to forecast the gross service in telecommunications,subjective guided data is introduced in consideration of the impact of industry convergence on future telecom industry.
出处
《吉林大学学报(信息科学版)》
CAS
2012年第6期598-603,共6页
Journal of Jilin University(Information Science Edition)
基金
国家自然科学基金资助项目(70473006)
关键词
小波神经网络
电信行业
参数初始化
经济预测
自适应学习速度
wavelet neural network
telecommunication industry
initialization of parameters
economics forecasting
adaptive learning rate