The plots shown are cumulative plots - clustering of these values may be significantly easier if you consider their individual values, not their cumulative values.
We can suggest many different clustering algorithms for these values, but with only one sample it is difficult to determine what methods will be most successful. Some questions to consider are:
- Do you know how any clusters you want?
- Do you know absolute thresholds for these clusters?
- Will the values always be arranged into these groups of similar values?
Clustering is highly dependent on the behavior of the data over the larger dataset that you would like to be clustering within.
Finally, without knowing the purpose of the data, we are not able to comment on appropriate segmentations of this data that would yield good results for your application. In some cases, the first singular value may be enough, while in others, throwing away any of the eigenvectors in your transformation could be detrimental.