Predicting the Upcoming Primaries Using Twitter Data (drive.google.com)

TL;DR "According to this model, we should expect Sanders to win California by a significant margin, roughly 150 delegates. Sanders should also win North Dakota and Montana, although the number of counted Twitter users is small enough that an upset wouldn’t be surprising. Clinton has a slight lead in New Jersey, while Sanders has a slight lead in New Mexico. Currently Sanders is behind by about 290 delegates, and could catch up by as much as 230 delegates during the June 7 primaries. While this is not enough to overtake Clinton in pledged delegates, it does approach the uncertainty estimates. An upset would be unlikely, but not completely out-of-the-question."