0
$\begingroup$

I have time series of cars' speed (speed observed over time for 50 different drivers driving the same car). I want to find the groups of "similar" driving styles based on these speeds. Therefore, I first apply DTW to compute a dissimilarity matrix between each pair of car speeds. And then use Hierarchical clustering using this matrix.

My question is: Are there any other clustering algorithms that I could use for this purpose? I have studied different algos. that were mentioned here, but the examples I find use non-time series data. Please direct me to any relevant resource.

You can see some sample data in this stackoverflow question. I use R.

$\endgroup$
1
$\begingroup$

You can use TADPOLE [a]

You CAN you k-means. The convergence in not guaranteed, but in practice it works.

However, I agree that with a small data-set, HAC makes sense

[a] Accelerating Dynamic Time Warping Clustering with a Novel Admissible Pruning Strategy http://www.cs.ucr.edu/~eamonn/Speeded%20Clustering%20Paper%20Camera%20Ready.pdf

$\endgroup$
1
$\begingroup$

Plenty.

Almost all except k-means and GMM can be used with distances such as DTW.

But many will "require" more than 50 samples to yield good results. With such tiny data, I would go with HAC.

$\endgroup$
  • $\begingroup$ Could you please name a few? $\endgroup$ – umair durrani Oct 4 '17 at 2:54
  • 1
    $\begingroup$ DBSCAN, AP, SC, canopy, ... but again, they work much better if you have enough samples. $\endgroup$ – Anony-Mousse Oct 4 '17 at 5:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.