Predicting television advertisement reach with machine learning models

Despite the entry of many media services, television remains the most used media service and accounts for the largest advertising spending globally. One of the main metrics for measuring the successfulness of a television advertising campaign is reach, the percentage of the intended target audience that has seen the television advertisement. To help plan television advertisements, the industry aims to find new methods for predicting television advertisement reach more accurately. Therefore, it is of interest to explore the possibility to utilize machine learning regression models. This report examines how well four machine learning regression models are suited for predicting reach based on historical campaign data. The results indicate that the best-performing model is an XGBoost model with a mean absolute percentage error just below 5%. The report also describes which features impact reach the most and if data augmentation can improve the performance of the machine learning models.