A Metadata Approach to Predicting Twitter User Geolocation

Abstract

As the volume of available data created from social media grows, creating value out of this information becomes an interesting challenge. This work presents an approach to identify a users country of origin from just looking at a single tweet. The approach focuses on using only metadata, while disregarding the actual content of the tweet. This work also looks at the scalability of the approach using Naive Bayes algorithm. The experiments concludes that using metadata to predict a user geolocation is a viable approach that scales to a stable accuracy with increasing data size.