I'm aware that passenger generation is in line for a shake-up in the next major release, but before this happens I thought it would be worth flagging some of the anomalies in current code, to see if these are the sorts of things that are in line for a fix.
In short, some areas seem to generate a disproportionate number of passengers for their size. These areas are sometimes (often?) on the outskirts of large cities.
https://dl.dropboxusercontent.com/u/61716/passengers.rar (https://dl.dropboxusercontent.com/u/61716/passengers.rar)
Fazakerley (in frame) seems to generates passengers several orders of magnitude above any nearby location. (An instructive comparison is St Helens Central or Widnes, to the east and southeast respectively, which generate *far* fewer despite seeming to have comparable populations.
These are also inconsistent: check Blake Street in north Birmingham. This gave a huge passenger flow for very small population in the previous month, but if you fast-forward you'll see that this tumbles in the current month.
The main striking contrast here is between the likes of Fazakerley & Blake Street and the likes of St Helens (other examples of low-flow urban stations: Bexleyheath, Welling and Barnehurst).
It may well be that there's nothing that can be done about this now because of upcoming changes, but I wanted the anomalies to at least be on the record.
Thank you for pointing this out. I strongly suspect that these anomalies are caused by the fixed distance ranges, which are removed in the passenger-generation branch: what I think probably happens is that particular towns happen to be in exactly the right place to be the mid-range destination of nearly everyone from a very large urban centre in circumstances where little else is in range and accessible by transport. The numbers of passengers "generated" in these towns includes the number of return trips, which is based on the number of incoming trips from other, larger towns.
Hopefully, this will be solved in the next major release. Thank you for the observation, though!
Thanks for the reply, James. I've long suspected that something like what you describe is going on. Given the values as I've set them, it seems that Fazakerley would be at about the minimum distance for a mid-range passenger from the West Midlands, so this may well explain it. And indeed, anecdotally there do seem to be a lot of passengers generated here heading to the West Midlands.
Nevertheless, there are also plenty of passengers generated here for the London area, whereas almost none are generated at St Helens. And after a few hours of the month have passed, even Fazakerley's generation slows down to a crawl. So maybe there is something else going on too. I guess we will have to revisit this after the upcoming changes to see if it goes away.
It will probably be easier to see whether there are any remaining anomalies after the new system is implemented, as the new system is really quite different in a multitude of respects from the current one.