Home > Datasets, what i've been doing > Community Clustering part 2

Community Clustering part 2

December 8th, 2010 Leave a comment Go to comments

I calculated the communities as described  in data-pipelines and from this generated two lists of nodes for every week in October, and for all weeks.  I took the largest list for each, and used that as the community for which to test the LBR algorithm.

updated-plot-comm

Updated plot, showing the delivery ratio for LBR-Community, one (dotted) line for the communities found in each week, and overall.

It seems that, as expected, routing within the community gives an advantage to the algorithm, and in fact, when the community knowledge is derived from the full dataset, then LBR performs better that the average of the rotated dataset. This plot is slightly misleading however, as the data for the previous algorithms is the rotated averages, whereas the data for LBR-Community is for straight runs. Below is the plot against straight runs of each other algorithm.

updated-plot-comm-straight

This plot shows straight runs of all algorithms, and the updated LBR-Community runs.

In this plot, it is obvious that the difference is less profound as with the average plots, however, I am considering running the rotation scheme again for LBR-Community.

  1. No comments yet.
  1. No trackbacks yet.