Which clustering method is suitable for which kind of data?

I would like to know

1. K-means is best suited for clustering of which type of data? 开发者_开发技巧
2. When k-means fails? for which type of data set k-means does not give accurate answer?
3. COBWEB is best suited for clustering of which type of data?
4. When COBWEB fails? for which type of data set COBWEB does not give accurate answer?

1)Looking at some Infinite training Finite training, I can say that K-means is best suited for any kind of data which can be divided in to vectors and best for quantitative data.

2)K-means fails When the numbers of data are not so many,When initial condition is sensitive or flickering ,which gives different results

K-means can have issues in high dimensions when using euclidian distance as everything ends up being "close".

What type of clustering are you trying to do?

I have had problems using K means clustering with a data set that included Dip and Dip Direction (points on the surface of a sphere). In the end I had to create a crude genetic algorithm to do the clustering.