#recalculate centroids of clusters
        for key in self.assignments:
            clust_sum=0
            for k in self.assignments[key]:
                clust_sum=clust_sum+data[k]
            self.centroids[key]=[m/len(self.assignments[key]) for m in clust_sum]-- error in this line,int is not iterable. 

change clust_sum declaration to np.zeros(len(data[0]))
change clust_sum=clust_sum+data[k] to clust_sum=np.add(clust_sum,data[k])

change it to following code

#recalculate centroids of clusters
for key in self.assignments:
clust_sum=np.zeros(len(data[0]))
for k in self.assignments[key]:
clust_sum=np.add(clust_sum,data[k])
self.centroids[key]=[m/len(self.assignments[key]) for m in clust_sum]

if closest_clust in assignments:
    assignments[closest_clust].append(ind)
else:
    assignments[closest_clust]=[]

this code is error! when the element match the clus , the element will not append to the clust ,it make the result error!the right code like this:

assignments.setdefault(closest_clust,[])
assignments[closest_clust].append(ind)

Hi Alex, Can you please tell about/ describe a bit of data. What number or the row suggest what.?

I see that tslearn.clustering natively supports dynamic time warping via TimeSeriesKMeans.

However, TimeSeriesKMeans is quite slow. I would like to use this implementation which from the code looks like it has more optimization via locality constraints. I'm not sure it's actually faster but I have my fingers crossed it is.

Can someone point me to how I can use @alexminnaar's implementation to output cluster labels per series?

I can see that this implementation robustly outputs the average cluster curves but I don't see it outputting the labels for the entire time series data.

I suspect it is in this block, but I'm having trouble parsing it.
if closest_clust in assignments: assignments[closest_clust].append(ind) else: assignments[closest_clust]=[]

Any pointers would be appreciated.

Thanks for the excellent ipython notebook.
I have a doubt, can you please clarify.
Using k-means clustering, what is the conclusion? there are no clusters as such.
How can we predict cluster number for a new input, based on such results?

Please revert back.

Hello,

For some reason I can't load your notebook on

http://nbviewer.ipython.org/

I've followed the exact method (for Kmean Clustering) you've written in the tutorial.
But it generates following error:

100
iteration 1
100
iteration 2
69
iteration 3
48
iteration 4
38
Traceback (most recent call last):
  File "spatio-time cluster.py", line 177, in <module>
    for i in centroids:
TypeError: 'NoneType' object is not iterable

The datasets are same as your Github repo. I am also attaching the code. Please let me know, where I am doing wrong.