I looked in the original data set [1] to see if this was in there; unless "SMS" meant "SMS search", it may not be at all. If it isn't in there it would be interesting to put it in the model as it stands now and see what the model thinks of the probability of shutdown.
