Estimating Turkish television ratings by artificial intelligence

In an environment where advertising budgets face downward pressure, the high cost of television advertising has led to increased scrutiny on media planning and buying, which is the most expensive part of advertising. Ad costs are directly linked to program ratings, as advertisers aim to reach a targeted gross rating point (GRP) during their campaign. Therefore, accurately predicting ratings before broadcasts is crucial for both broadcasters and advertisers. The main objective of this study is to predict Turkish television ratings using artificial intelligence. Within this scope, models have been developed to predict future ratings based on historical data using various algorithms. This allows for the evaluation of the success of artificial intelligence algorithms in predicting television ratings. The research also aims to determine the accuracy rate of AI models in rating prediction. In the study, program ratings during the PT1 time slot of the seven most-watched channels in Istanbul, Ankara, and Izmir between 2013 and 2021 were analyzed, along with data on channel, program genre, program repetition status, program day, program duration, and temperature. Statistical tests, correlation and regression analyses, and artificial intelligence models were employed. The results of the research demonstrate that considering temperature can enhance the accuracy rates of rating predictions for television channels, advertisers, and ad agencies. Furthermore, the use of Decision Tree and H2O AutoML algorithms in this process has been observed to improve the accuracy of predictions. These findings highlight the effectiveness of artificial intelligence and machine learning techniques as valuable tools in rating prediction, playing a significant role in achieving accurate predictions. Channels, advertisers, and ad agencies can obtain more precise and reliable rating predictions by utilizing these algorithms.