I have a CSV file of the latitude and longitude coordinates of trains in the format [timestamp, runID, latitude, longitude]. The problem is that I have tracked multiple trains over the course of a few days, and once a train is done running it will then get a new runID for the next service it runs, meaning that some runs will start at a timestamp other than 0, and a different location.

How should I represent this data, and what predictive models would best suit such a dataset?

  • $\begingroup$ This doesn't make sense. You told us the 1st field was timestamp, which I take to be an ISO-8601 UTC instant in time, often represented as seconds since midnight 1970-01-01Z. You said that "some runs will start at a timestamp other than 0" -- are we to understand that most trains started their run at midnight 1970-01-01Z ?!? $\endgroup$
    – J_H
    Feb 9 at 0:43
  • $\begingroup$ No, the timestamp is seconds since the recording started. $\endgroup$ Feb 9 at 0:44
  • $\begingroup$ Ok, that sounds more plausible. But I'm still astonished. You have a large number of train runs which all started at timestamp 0, so for example they all simultaneously left their various rail stations at noon UTC on 2024-02-07 ?!? That still makes little sense. The various (lat, long)'s are universally interpretable. But the timestamps sound like they are not. Begin by making them proper ISO-8601 UTC timestamps. Then you can worry about relative time offsets or crow-flies-distance position offsets for a given train. $\endgroup$
    – J_H
    Feb 9 at 1:18
  • $\begingroup$ Adding a sample dataset might be beneficial for us. $\endgroup$
    – m13op22
    Feb 12 at 16:13


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.