On Quantifying Changes in Temporally Evolving Dataset

In this paper, we present a general framework to quantify changes in temporally evolving
data. We focus on changes that materialize due to evolution and interactions of features ex-
tracted from the data. The changes are captured by the following key transformations: create,
merge, split, continue, and cease. These transformations have been used successfully to study
the temporal dimension in various domains like text, scientific data, bioinformatics and so-
cial networks. Once discovered, the user has to sift through the discovered transformations to
find useful ones. This is a cumbersome and error prone approach. Moreover, the number of
discovered events can be large (e.g. 100000 events in DBLP) which renders the manual pro-
cess infeasible. Towards this goal, we present a general algorithm to automatically rank/score
these transformation. First, we identify various factors which influence the importance of each
transformation. These factors are then combined using a weight vector. The weight vector
encapsulates domain knowledge. We evaluated our algorithms on various real data sets arising
from diverse domains. In this paper, we present our results from the following datsets: DBLP,
IMDB, Text and Scientific Dataset.