Neural Gas Cluster Report Incomplete

[dgibson]

Hi, I am attempting to use the K-Centroids Cluster Analysis tool to group customer locations based on their geographic location. I have successfully used the K-means option and have found it to be very useful. Out of interest I switched to the Neural Gas method to compare results. Although the neural-gas clusters seem to be more appropriate, the report generated on the R side of the tool is missing clusters. If I request 70 clusters for example, 70 clusters are presented in section 7 of the report output but only 57 are shown in section 5 (where the average size is shown). Equally, when I use the Append cluster tool, only 57 clusters are given. 

As I'm new to the Neural Gas method I'm a little confused but it looks to me like some of the clusters (and some of the customer locations) are missing when using neural-gas. 

Any tips anyone can give are appreciated. 

[joshuaburkhow]

Since you are working with Geographic location data maybe this is better handled using the spatial tools. They are very powerful and since you said you are grouping based on geolocation this seems to be the track you want. 

Are you able to share your workflow? Maybe we can help to investigate why it's picking up the 57 and not 70. 

[dgibson]

Thank you Joshua for the suggestion. My reason for using the K-means clustering tool is the size of the data set. I'm using data from all states of Australia and want a way of grouping customers for transportation purposes. I'm interested in their proximity to each other rather than arbitrary groupings like post-codes. Clustering takes the hard work out because I don't need to determine starting points and try all combinations myself. 

Interestingly, I've noticed the number of clusters returned by Neural Gas is always less than requested. For example, if I requested 50 clusters around 40 will be returned whereas if I request 70 for the same data set, 57 will be returned. 

A screenshot of my workflow is attached.