Is it unusual to use cluster analysis algorithms such as k-means or average linkage clustering for a single vector, i.e., univariate datasets, but, it is incorrect?
there is nothing wrong to apply cluster analysis to univariate data
you can even hunt for the optimal k-mean solution via dynamic programming !
Article Ckmeans.1d.dp: Optimal k-means Clustering in One Dimension b...
Not necessary. For instance, In subspace clustering, samples are represented by a linear combination of other samples. Emphasis is on samples, not the feature space.
There are some general statistical density functions or models which play a general role for data analysis. Such models can be employed as sing to multi variables without any issue. See e.g., Gaussian distribution, Von-Misses Fisher distribution, Kent Distribution and Bingham distribution etc.
Thanks to everyone for your answers, as a conclusion of the current discussion, formally is possible the use cluster analysis in a univariate space, however, some contributions are more specific for this case, for example, Ckmeans.1d.dp: Optimal k-means Clustering in One Dimension by Dynamic Programming by Haizhou Wang and Mingzhou Song. Now I am working with this packed and the outputs look well and more consistent. Regards and thanks again
- Badges
- Science topic
- Similar topics
- Vectorization