基于ARIMA与BP神经网络的收视率组合预测模型  

Combined Prediction Model about Audience Rating Based on ARIMA and BP Neutral Networks

作　　者：黄玲莉[1] 刘小龙[1] 

机构地区：[1]华南理工大学工商管理学院,广东广州510640

出　　处：《电视技术》2015年第9期117-121,共5页Video Engineering

基　　金：国家自然科学基金项目(71071057);中央高校基本科研业务费专项资金项目(2014XMS04)

摘　　要：针对收视率数据的非线性、突变性等特征,仅采纳单一的预测方法不能全面描述收视率的变化规律,因此提出了一种组合预测模型（ARIMA-BP）.首先采用自回归移动平均模型（ARIMA）对收视率进行预测,提取收视率的线性变化规律,再利用BP神经网络对ARIMA模型的预测值进行进一步的组合预测,提取收视率的非线性变化特征.本文以广州电视综合频道及在广州市场上的其余8个电视频道19：00 ∽20：00时间段的收视率为例进行实证分析,结果表明组合模型比单一采用ARIMA、BP神经网络进行预测的拟合效果好、精度高.The audience rating is nonlinear and salutatory, unitary method cannot predict its change regulation accurately, for these, a combined prediction model （ARIMA--BP）is put forward in this paper. Firstly, using the ARIMA model to predict its change regulation of liner section. Secondly, using BP nentral networks to predict its regulation of nonlinear section and then, combining the above two resuhs. This paper uses the comprehensive channel of GZTV and other eight channels in Guangzhou nmrket during 19：00 -20：00 as examples to analyze, the result indicates the fitting effect is good enough.

关 键 词：ARIMA模型 BP神经网络 电视收视率 预测 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]