I think there is a bug during the calculation of the scaling factor. If you use the following formula, the scaling factor will be irrelevant to the value of beta.

scaling_factor = (1 - self.beta) * scaling_factor + (self.beta) * self.scaling_factor

Can you please advise me how to fix this bug? Thanks!

I am using a table with lat lon and weights with ~10.000 entries and I would like to create 10 clusters. The maximum weight is = 150.000. It contains few (~20) zero weights.

WKmeans always throws the same error:

ValueError: One or more clusters disappeared because all points rushed away to other cluster(s). Try increasing the stickiness parameter (beta).

Changing the beta or alpha doesn't change anything. Instead I noticed, when I divided my weights by let's say 1000 it worked again.

This led me to the function causing the error:

    def _has_converged(self):
        """Check if the items in clusters have stabilised between two runs.

        This checks to see if the distance between the centroids is lower than
        a fixed constant.
        """
        diff = 1000
        if self.clusters:
            for clu in self.clusters:
                # For each clusters, check the length. If zero, we have a
                # problem, we have lost clusters.
                if len(clu) is 0:
                    raise ValueError('One or more clusters disappeared because'
                                     ' all points rushed away to other'
                                     ' cluster(s). Try increasing the'
                                     ' stickiness parameter (beta).')

It seems that diff = 1000, is causing the issue as it's probably not suited to larger numbers or a large difference between min an max weight as in my case.

Could anyone recommend a dynamic value generation instead of the static value? Can't it just be like max weight / some factor?

First of all: great repo!
Not an issue but rather some advice needed. When passing my data I would like to pass an id for every coordinate in order to process the data afterwards.

Assume this data where lat, lon are the coordinates (X) and weight is the variable c. My aim is to fill the column cluster which I would normally do via a simple pandas merge command, pretty much like in this medium article for sklearn's kmeans (see step 6).

As your wkmeans method doesn't have the function predicted_kmeans = kmeans.predict(X, sample_weight = Y) I was wondering what the easiest way was to fill the cluster column.

My main idea was to create a fake id by concatting lat+lon but this would lead to issues with duplicates.

I tried creating a 3D-array with lat, lon and id but it returned - for my use-case - "false" 3D-results.

Currently, I'm just passing the data from a pandas df.

X = np.array(df[["lon","lat"]].astype(float))
c = np.array(df['weight'].astype(float)) 

Any advice?

Thanks in advance!

I tried all the same python package version but it does not run.