Coupling machine learning and integer programming for optimal TV promo scheduling

Abstract

Optimal TV promo Scheduling is the process of scheduling promos over breaks in order to maximize promos and programs’ viewership. It is a complex task to tackle since viewership is an uncertain quantity to estimate affected by uncontrollable events, many business requirements need to be satisfied and unexpected events may require the definition of a new schedule in a very short time. In this work, a new efficient framework for solving the Optimal TV promo Scheduling problem is introduced by formulating the problem as an integer optimization problem where the viewership is estimated through Machine Learning models. Different objective functions are defined and benchmarked. Numerical results on real word instances show the effectiveness of the resulting framework in solving the Optimal TV promo Scheduling problem in a very short amount of time leading to good or optimal solutions and improving schedules KPI provided by business experts.