Focus on single AP or set of APs

[mentat]

It would be way cool to be able to select a single AP and be able to display all the datapoints related to it colored by signal level.

[uhtu]

iiinteresting. it would require online access to wigle.net or the dataset would be
*HUGE*

[mentat]

Very nice implementation! I love it!

[mentat]

It also becomes fairly obvious if there's bogus data. What kind of data clean up should we do?

[uhtu]

its a hard problem to automate; as more powerful cards and directional antennae, low quality gps gear, etc. will appear to skew the results for an otherwise valid ap. this will correct itself over time as an ap is reobserved, and its average is improved. just using standard deviations to exclude an outlying point (the 'right'ish way) will exclude valid observations as well. once an ap has sufficient coverage (yaay heuristics!) throwing out bad points gets alot easier.

[mentat]

There appear to be some points for which the distant (invalid) point is over 100 miles away. It should be possible to do something about that. It's not the center plotting that I'm worried about (though the center of above networks is obviously wrong). It's the polygon drawing that really gets disturbs (and can't necessarily get correct with more observations). My math was never very good, but it should be possible, given FCC limits and the over all shape of the network (directional, omnidirection, or "panel" type patterns), to make a good guess at what's even possible.

[mentat]

could also be taken into account using dates.