基于TEI@I方法论的中国季播电视综艺节目收视率预测

Forecasting audience ratings of China's seasonal entertainment TV shows based on TEI@I methodology

本文以TEI@I方法论为指导,提出了一个季播电视综艺节目收视率预测的研究框架.季播电视综艺节目是中国电视行业近三年发展的新兴趋势,收视率预测研究对于其排编优化和广告资源科学定价具有重要的指导意义.本文在传统数据的基础上,加入了百度指数和新浪微指数,通过建立线性回归模型发现如下规律:首期收视率对后期收视率具有锚定作用;平均收视率呈现逐年下降趋势;每年冬季和每周周五易出现收视高峰;百度指数和新浪微指数与收视率存在显著正相关.除了线性回归模型外,本文还建立了RBF神经网络、支持向量回归模型,并进行了模型集成预测,实证结果表明:加入百度指数和新浪微指数能够提高预测精度,集成模型比单一模型更能有效地预测节目收视率的走势.

This paper proposes a research framework for forecasting audience ratings of seasonal entertainment TV shows based on the TEI@I methodology. Seasonal entertainment shows are emerging in China＇s television industry in the past three years. The forecast of their audience ratings is fundamental to the optimization of program scheduling and the scientific pricing of advertisements. As supplement to conventional data, the Baidu index and the Sina microblog index have been used in forecasting audience ratings. The resulting linear regression model demonstrates the following conclusions：the ratings of the first episode dictate those of the subsequent episodes; the average ratings are decreasing over the years; ratings tend to peak in winters and on Fridays; the two Internet indices show significant positive correlation with the ratings. Besides linear regression, RBF neural network, and support vector regression are experimented, and the integrated technology are employed. The experiment results show that the utilization of big data on internet make noticeable improvements to the forecasting accuracies, and the integrated technology outperforms in predicting the future trend of ratings.