Andrew Gray (![[info]](http://l-stat.livejournal.com/img/userinfo.gif){kind=small} shimgray) wrote,@ 2009-07-08 22:16:00 |
|
Mapumental
So, I've spent some time over the last week playing with Mapumental, a more developed version of these travel-time maps which mySociety produced in 2007; they've a blog post about the new site here.
Essentially, it's a graphical frontend to a database of public-transport travel times; you feed in a postcode, and it displays the time to get there from any other point as an overlay on a map. This is then tied into two other datasets; house-price values and "prettiness", so you can filter the result by the amount you are willing to pay, and how "nice" an area is to be in. There's a video demonstration of it here:
Despite the appearance of that video, it works throughout the UK (the mainland, at least, I've not tried NI).
It's pretty effective. Their travel-time data, for the only commutes I know well, can be about five to ten minutes off, but this is probably an effect of exact bus times and the estimated walking time, so it's not much of an error. The "scenicness" slider is a bit awkward, but once the data's scaled properly it'll no doubt work better.
Unfortunately, real life has been very hectic, so I've not had much time to study it as I'd like; I've tried it for a few places I know, and not much else. Some thoughts, in no particular order, on what it's currently missing:
I really do look forward to seeing where this goes next. It's a great toy, and potentially really useful for helping make quite complex decisions.
So, I've spent some time over the last week playing with Mapumental, a more developed version of these travel-time maps which mySociety produced in 2007; they've a blog post about the new site here.
Essentially, it's a graphical frontend to a database of public-transport travel times; you feed in a postcode, and it displays the time to get there from any other point as an overlay on a map. This is then tied into two other datasets; house-price values and "prettiness", so you can filter the result by the amount you are willing to pay, and how "nice" an area is to be in. There's a video demonstration of it here:
Despite the appearance of that video, it works throughout the UK (the mainland, at least, I've not tried NI).
It's pretty effective. Their travel-time data, for the only commutes I know well, can be about five to ten minutes off, but this is probably an effect of exact bus times and the estimated walking time, so it's not much of an error. The "scenicness" slider is a bit awkward, but once the data's scaled properly it'll no doubt work better.
Unfortunately, real life has been very hectic, so I've not had much time to study it as I'd like; I've tried it for a few places I know, and not much else. Some thoughts, in no particular order, on what it's currently missing:
- The model only generates travel plans for "arriving at point X at 9.00am". This is fair enough for commuting, but it'd be nice to see it able to shift around that time a bit - or to handle shiftworkers.
- Some way of feeding in two destinations - for a couple who work at different places, for example, someone with a multiple-site job, or a job and a school for the same family.
- A model for car travel, or bicycle - at the moment it's solely public transport. The datasets are certainly out there, but I don't know how easy they are to apply.
- The geographical model is a little shaky. Where it doesn't have transport beyond a point, it just gradually grows a circle around it to represent walking, regardless of what it's growing over. The result is that some of my recommended areas around here involve walking across the Thames, or over motorways... and near the coast, you have the weird effect of anywhere a few hundred metres offshore being acceptable.
I really do look forward to seeing where this goes next. It's a great toy, and potentially really useful for helping make quite complex decisions.
