A Comparative Study on the Time Series Models for Forecasting Facebook Reactions

Abstract

The Facebook reactions were used over 300 billion times during their first year of existence. Research on reaction activity is essential especially for the digital marketing purpose. The market needs to understand how Facebook reactions fluctuate to forecast the best period to post advertisements on Facebook that yields the highest number of reactions. In this study, several time-series models are used to forecast the number of Facebook reactions over a certain period for different domains. A comparative study is done to evaluate the performance of each model, in terms of strengths and weaknesses. Keywords: Forecasting, Facebook reactions, time series model, ARIMA, SARIMA