Game of Shows: Using linear regression and integer programming to optimize television ratings

Abstract

The rapidly changing landscape of the television industry has motivated researchers to find ways to optimize Nielsen ratings, and thus maximize revenue. While other papers in this field focus on optimal scheduling, this thesis focuses on how content and other inherent attributes of a broadcast television show can be chosen to increase viewer ratings. A two-step linear regression is used to determine which attributes significantly contribute to ratings, and then to generate a synthetic data set. A modified integer program is then run on the synthetic data to select maximizing portfolios of shows for each broadcast network. The regressions found the presence of an experienced show developer, the presence of related content, and average character age to be significant attributes. The integer program was then able to exceed current P18-49 ratings for each network by an average of 0.17 percentage points.