Clustering for trajectories

I have a large number of temporary lat / lon.

I am trying to find k-clusters of paths from this data. What is the best approach for this?

Thank.

Edit:

How do I create functions for my data (lat / lon + time) to use kmeans / hierarchical clustering?

Edit:

I hope this makes it easier

Here is an example of what my data looks like:

Trajectory 1:

lat1, lon1 at time1
lat2, lon2 at time2
...
lat55, lon55 at time55
Trajectory 2:

lat343, lon343 at time343
lat344, lon344 at time344
...
lat376, lon376 at time376

And again and again (a couple of additional trajectories).

So to speak, I have 200 of these trajectories, I want to group them into 2 groups. How do I approach this?

Should I use kmeans / HAC for this, or should I look at another method?

Edit:

, k , k .

. .

, , - :

Direction 1:
Trajectory 4
Trajectory 5
Trajectory 7
Direction 2:
Trajectory 44
Trajectory 2
Trajectory 27

...

Direction 10:
Trajectory 17
Trajectory 8

. - ( ), - .
: lat/lon - , .

. ?

Edit:

( ):

Trajectories and end result

.

+5
3

K- .

, , 180. , .

, , . , k-.

. , , , , appopriate . , , . , !

+3
+1

(DTW) , ( , ). DTW .

For your dataset, I would extract the orientation sequences only because including the distance component can cause problems if the trajectories are traversed at different speeds or if samples are taken at heterogeneous time intervals.

0
source

All Articles