Predicting television programs success using machine learning techniques

Abstract

In the ever-evolving media landscape, television (TV) remains a coveted platform, compelling industry players to innovate amid intense competition. This study focuses on leveraging machine learning regression models to precisely predict TV program reach. Our objective is to assess the models' efficacy, revealing a standout performer with a mean absolute percent error of just under 8%. Significantly, we identify features exerting a substantial impact on predictions and explore the potential for model enhancement through expanded datasets. This research extends beyond statistical insights, offering actionable implications for TV channel managers. Empowered by these findings, managers can make informed decisions in program planning and scheduling, optimizing viewer engagement. The temporal analysis of evolving trends over time adds a nuanced layer to our study, aligning it with the dynamic nature of the media landscape. As television retains its dynamic force, our insights contribute not only to academic discourse but also provide practical guidance, enhancing the competitive edge of television channels.