Title:Using Social Media to Predict the Future: A Systematic Literature Review

Abstract:Social media (SM) data provides a vast record of humanity's everyday thoughts, feelings, and actions at a resolution previously unimaginable. Because user behavior on SM is a reflection of events in the real world, researchers have realized they can use SM in order to forecast, making predictions about the future. The advantage of SM data is its relative ease of acquisition, large quantity, and ability to capture socially relevant information, which may be difficult to gather from other data sources. Promising results exist across a wide variety of domains, but one will find little consensus regarding best practices in either methodology or evaluation. In this systematic review, we examine relevant literature over the past decade, tabulate mixed results across a number of scientific disciplines, and identify common pitfalls and best practices. We find that SM forecasting is limited by data biases, noisy data, lack of generalizable results, a lack of domain-specific theory, and underlying complexity in many prediction tasks. But despite these shortcomings, recurring findings and promising results continue to galvanize researchers and demand continued investigation. Based on the existing literature, we identify research practices which lead to success, citing specific examples in each case and making recommendations for best practices. These recommendations will help researchers take advantage of the exciting possibilities offered by SM platforms.