Abstract:
TV rating is a numeric estimate of the popularity of TV programs in households with a
television. Investment planning of television media, which is the biggest shareholder of
media investments, is made according to TV ratings. The aim of this study is to develop
a machine learning model that can predict the TV ratings of Turkish TV series in a
practical manner. Better forecast of TV ratings has the potential to reduce the risks of TV
investments. In this context, four prediction models were developed by using machine
learning techniques for TV series in the prime-time broadcast between 2014-2018 in this
study. The main distinction between these models is the attributes that represent a TV
program. The attributes that are used in this study can be categorized as: time-based,
episode-based, series-based, and program impact on social media-based (represented
using Google Trends) attributes. In the experiments, a theoretical forecast performance is
established first by using time-based attributes that factors in the future changes in the
TV program, which is then used as a baseline for comparison with practical forecast
models. The experiments show that, the proposed models can achieve up to 1.46% error
rate for theoretical models and 6.6% error rate for practical models.
