Home > Datasets, what i've been doing > Rotating the data

Rotating the data

November 25th, 2010 Leave a comment Go to comments

Met with Pádraig and Davide briefly to discuss the poor performance of the LBR algorithm. Pádraig had previously suggested that we run multiple simulations over different time slots, and Davide and I had talked about and planned to do split the runs up as follows:

timerotation

Measuring the delivery ratio, and cost for PBR, LBR, Prophet, Random 1.0, Random 0.2,  Random 0.0 and Flooding, will give us a good picture about the nature of the data, and we will be able to see if the previous plots are representative of the data.

I will also try to adapt the ranking mechanism, to include only the top (most seen) 10% or 20% of cell towers, to see if that makes any difference to the results.

Update:
Using the method above, I ran the simulations for LBR, PBR, Unlimited Flood, Random 1.0,0.0,0.2 and Prophet. Below is the average delivery ratio over 18 runs, for each  protocol in 28 day time windows, at 7 day intervals,  between  27 Aug 2004 and 21 Jan 2005.

Average delivery ratio over 18 runs for each protocol, between Aug 2004 and Jan 2005

There does seem however, to be a few issues with the number of lines returned in the logs, this suggests that somehow, I have mis-calculated, probably just for the end  time period used, the start and end time period, but this seems to only account for the last few hours of each run (out of the month). So until I find the error, I think these results are roughly correct.

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