ABSTRACT
One consideration for social network providers is that this data will accumulate over time. Because of the highly desirable nature of social network longitudinal data, it is unlikely that a network such as Facebook will ever delete data they obtain. Even if later revisions directly contradict previously entered data, this is still valuable in tracking change in attitudes over time. Because of this, performing calculations and inference on this data will become more difficult over time, because of the size of this data. However, we now show that by intelligently examining this data, we are able to partition the full data into smaller sets, which both increases classification accuracy and reduces the time required for classification.
