Yeah, I agree, it’s tedious, and should be better.
Currently part of the problem is that the processing engine treats differently bogus segments that were auto detected and bogus segments that we’ve manually marked as such. The reason it treats them differently is auto detecting bogus is very inaccurate, so I basically just don’t trust it to be acted on without human intervention. Reliably detecting bogus location data is a hard problem!
If all samples/segments within an item are manually marked as bogus, then the processing engine will merge that item into one of the surrounding items. But if some of the samples/segments within the item are only automatically marked bogus, it won’t do that merge. Which is partly why you see auto bogus items hanging around in your timeline, leaving you with more cleanup work to do.
But I’ve got it on my todos list to change that - to make it so that the processing engine treats auto bogus the same as manual bogus. I figure… how bad could it be? So let’s find out! There’s a chance it could work well enough (even if auto detecting bogus is still very unreliable and inaccurate), which would mean less cleanup work.
Trust Factor notes
One other thing to try is to try marking more samples/segments as stationary, instead of bogus, if they’re near enough. If they’re marked stationary, then that trains the Trust Factor system to recognise common areas of drifting location data, which then allows it to reduce that drift by putting less trust in iOS’s reported accuracy levels.
By doing that, you can reduce the amount of bogus location data. It’s only really worthwhile though if the drifting data is within maybe 100 metres of the real location (or less, or more, depending on what kind area it is - you kind of need to just feel it out).
The “clean up” button for visits will also do this for you - marking the samples within the place radius as stationary, and those outside the place radius as bogus. Though sometimes I still go in and manually changing them between stationary<->bogus, based on my own human intuition, which [for now] is still better than that of the robots/machines.
LocoKit2
You might have seen me mention Arc Timeline Recorder or Arc Timeline Editor in some other discussions here. I’m currently in the process of rebuilding most of Arc and its underlying recording engine from the ground up. This allows me to get rid of almost a decade of cruft, bugs, and various decisions that made sense at the time, but with newer technologies no longer make sense.
Arc Timeline Recorder is the first app to use the new LocoKit2 recording engine, and is ready to go live on the App Store now. I’m just in the process of satisfying Apple’s App Store review team, so that they’ll put it live. Hopefully tomorrow!
Anyway, the above mentioned change to bogus handling is going to be part of LocoKit2. It isn’t yet, but that (and Trust Factor) are two areas I’m hoping to improve significantly in the rebuild. Fingers crossed it all goes well!
