Predicting TV Viewership with Regression Models

Abstract

In this paper were analyzed the applicability of regression models to the problem of real time viewership prediction. In this paper the analysis is based on precise viewership data obtained from the national company responsible for maintaining the broadcasting infrastructure. The viewership data and the TV schedule archive data were preprocessed to create a dataset on which the analysis is performed. Analysis was performed on the correlation of the number of viewers with time series trends and TV schedule features. The results were compared with several implemented and trained models, and compared with the performance of various tested models. In this paper is described the methodology of building the models and discussed the obtained results.