Faulty phase1 dataset?

Dear team,

while investigating the dataset I came across some objects where the entries appear faulty. For example, object 124 looks as follows:

ObjectID TimeIndex Direction Node Type
124 0 EW SS NK
124 0 NS SS NK
124 366 EW AD NK
124 1199 EW AD NK
124 1725 EW AD NK
124 2171 ES ES ES

Why is the EW-AD-NK node listed three times? This is a problem because even if I predict the entire object correctly, my reduced output will look like this:

ObjectID TimeIndex Direction Node Type
124 0 EW SS NK
124 0 NS SS NK
124 366 EW AD NK
124 2171 ES ES ES

Which, given the evaluation code provided, will yield a F2 of 0.71

Or do I have some fundamental misunderstanding of how the data is meant to be interpreted?

Best,
David

I don’t think it ia a mistake. It means the satellite is changing the pattern of life three times in the study period in the EW direction for Adjust Drift (AD) with Not Knwon propulsion. I think the objective is to detect those three nodes.

You can imagine them as it the satellites is drifting in longitud initially at 5deg/day, then 10deg per day and then 20deg/day, for example

Thank you for your reply! After looking at the papers again it seems like your understanding is correct. I was misled by the baseline ML solution, where the data pre- and postprocessing completely eliminates the possibility of detecting the same node twice.