The app shows my transport as being in an airplane but it was really a motorcycle. Is there any way to adjust this or for me to teach the app? And I went 15.6 km but it only shows a small fraction of that as defined modes of transport…Cheers.
Hi @Mavec!
Go to the timeline view for that day, then scroll down to find the timeline item that’s been incorrectly classified as airplane. Tap on it to get to the details view, then tap the EDIT button at top right, then choose motorcycle from the list.
That will then trigger an update of the activity summaries for that day.
1 Like
Thanks for your fast reply @matt , much appreciated. Will the app then intelligently learn? For example, I see today it has classified my motorcycle ride as a car. I’ve edited it to motorcycle but wonder if I will need to review activity every day…
1 Like
Yes absolutely 
Arc learns from every confirmation/correction you make, and will be then be less likely to make the same mistakes again. Your confirmations/corrections also help to make the app become progressively more confident in its own decisions, so that it will have less “unconfirmed” items to ask you about each day.
Some activity types are intrinsically hard to be certain about, so Arc will still often have trips/visits marked as unconfirmed (meaning that Arc doesn’t have enough confidence in its decision, and wants confirmation from you). For example things like car vs taxi vs bus, where they all travel on the same roads and at similar speeds.
But based on your feedback (confirms/corrects) Arc will still improve its confidence on even the tricky ones, by learning your personal habits. For example if you always take the bus to work but take the train home, then the ML (Machine Learning) models will learn those routes and time of day patterns, and be increasingly confident in its detection of those trips, and likely to stop asking you to confirm them.
1 Like
To add: You will also see this same learning in the detection of visits to places.
For example, I visit a small local mall every day, where the shops are all close enough together that they can’t be distinguished by location data alone. In the mornings I go to the cafe and stay about 1.5 hours, then in the afternoons I go to the convenience store for 5 minutes. Arc has learnt those patterns, so it knows that the morning visit is almost certainly the cafe, and the afternoon visit the convenience store, based on time of day and duration of visit.
So even though in terms of location data those two visits are basically identical, Arc has learnt to get them right on its own, and not to bother with asking me to confirm anymore. The magic of Machine Learning 
1 Like
Hey there 
just wanted to +1 this. For me, driving short distances across essentially flat land shows an altitude change of 5k+ feet 
which could be the culprit. I’ll keep editing for now but thought I’d add that data point. Thanks!
That’s a weird one! Are you going through any tunnels perhaps? iPhones produce altitude data by combining two sources: 1) GPS / GNSS data (GPS provides altitude as well as latitude and longitude!), 2) the phone’s barometer. GPS / GNSS altitude data accuracy tends to be slightly less accurate than lat/long (so from maybe 10 to 100 metres accurate), while barometer data can detect altitude changes down to centimetre accuracy. So the two work together to produce a composite of relative and absolute values.
Dramatic changes in altitude could be the phone losing GPS line of sight, combined with a change in air pressure - so basically a tunnel is the most likely culprit!
Arc filters the raw altitude data the same as it does raw lat/long data, so short bursts of nonsense altitude data won’t cause a mess. But a longer tunnel could create the situation where the incoming data begins to appear correct, given its consistency, and at that point the filtering won’t be able to recognise it as nonsense data.
Interesting! Thanks for the info. No tunnel, but I do cross a bridge over a river. Not a huge deal, just wanted to let you know. Only a few days into using the app, pretty neat so far!
1 Like
Hm. Curious. Wonder what the anomaly was then! If it keeps happening, let me know. There’s likely nothing I can do about it (it’s up to the phone to figure that stuff out), but I’m always curious about the different ways the phone acts under different conditions.