Success prediction of online news about TV series with machine learning, Google Analytics, and Twitter

Abstract

Journalism has adapted to the digital environment using web analytics and trend analysis to measure the success of its content. To optimize resources and increase visibility, new information needs arise in the editorial process. Therefore, this study proposes a cybermetric methodology that employs machine learning to predict the success of online news about television series, a growing theme whose virality is closely related to social networks. The methodology design consists of selecting indicators and tools, data collection, multiple linear regressions to predict success indicators, and validating prediction equations to obtain their accuracy. Prediction equations of success indicators have been obtained using an online media outlet as a use case, segmenting the data into three sets: all articles, TV series articles, and trailer articles. Validation has allowed for the comparison of equations and the selection of the most accurate equation. This research provides a tool that can be integrated into the editorial process to optimize its strategy, and it is a starting point for future research to improve accuracy in multiple ways.