
139Tweeque
Thus, at each step of the agglomeration, we choose the two concept clusters with maximum value
of the objective function Θ. If the score of the combined bag of concepts crosses the LCT, we return
the bag of concepts as the possible location vector and terminate.
To understand how agglomerative clustering basically works, consider a scenario in which the
location prediction algorithm returns an array of location concepts including (Los Angeles, 0.34),
(Long Beach, 0.05), (Irvine, 0.17), and many other concepts. Suppose the LCT for the algorithm to
return a cluster of concepts is 0.5. Then, if we simply combine location concepts on the ...