基于RBF神经网络的电视收视率预测  

The Audience Rating Prediction Based on RBF Neural Networks

作　　者：梁招娣[1] 刘小龙[1] 

机构地区：[1]华南理工大学工商管理学院,广州510640

出　　处：《河南科学》2013年第9期1428-1431,共4页Henan Science

基　　金：国家自然科学基金资助项目(50675069;71071059)

摘　　要：收视率受各种因素的影响,由于各因素间存在的复杂的非线性关系,传统的预测方法已不能反映收视率的非线性变化规律.为了提高预测的精度,提出了一种RBF神经网络预测方法.通过对收视率样本数据进行非线性定阶,对模型进行重构,然后利用RBF神经网络寻找最优参数并进行预测.以广州经济频道及在广州市场上的其他4个电视频道19:00—20:00时间段的收视率为例进行仿真分析,结果表明模型的拟合效果较ARIMA及BP神经网络的结果好.Audience rating is influenced by different factors. Because of the complex relatinships among different factors, traditional methods can not reflect its nonlinear change rules. In order to improve its precision, this study presents a forecasting method of RBF neural network. Through the nonlinear ordering rating data and reconstructing the model,the best parameters are found and forecast is done. The simulation analysis of audience ratings of Guangzhou Jingji channel and other four channels during 19：00-20：00 is carried out,the results reflect that RBF model is better than ARIMA and BP.

关 键 词：RBF神经网络 收视率 预测 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]