Analysis of home location estimation with iteration on Twitter following relationship

Abstract

User's home locations are used by numerous social media applications, such as social media analysis. However, since the user's home location is not generally open to the public, many researchers have been attempting to develop a more accurate home location estimation. A social network that expresses relationships between users is used to estimate the users' home locations. The network-based home location estimation method with iteration, which propagates the estimated locations, is used to estimate more users' home locations. In this study, we analyze the function of network-based home location estimation with iteration while using the social network based on following relationships on Twitter. The results indicate that the function that selects the most frequent location among the friends' location has the best accuracy. Our analysis also shows that the 88% of users, who are in the social network based on following relationships, has at least one correct home location within one-hop (friends and friends of friends). According to this characteristic of the social network, we indicate that twice is sufficient for iteration.

Home location prediction with telecom data: Benchmarking heuristics with a predictive modelling approach.

“…These survey methods have major shortcomings such as small sample rates, short survey durations, under-reporting and a high cost (Calabrese et al, 2011). Meanwhile, a variety of other data sources has been used, such as circulating bank notes (Brockmann et al, 2006), Foursquare check-in data (Noulas et al, 2012), tweets (Hawelka et al, 2014;Mahmud et al, 2014;Hironaka et al, 2016) and GPS data (Vazquez-Prokopec et al, 2013;Tang et al, 2015).…”

Quantifying uncertainty in population weighting of Twitter analyses for urban risk assessment

“…Geotagging tweets that are lacking geographic coordinates has been the subject of several scholarly studies (Chandra et al 2011, Hironaka et al 2016, Mahmud et al 2012, Hecht et al 2011, Wing et al 2011, Schulz et al 2013, Roller et al 2012, Compton et al 2015, Poulston et al 2017. These methods range from using screennames of followers to in-text interactions between users ("mentions"), language analysis, profile information analysis, and external text corpi to identify the location from which the tweet was generated from and/or the home location of the user.…”