I have a task to perform unsupervised cluster analysis on mixed datatypes: images, physical and business measures – continuous and categorical. Businesswise: there are images of products and their characteristics, when the task is to identify products that form coherent groups. I will appreciate links to papers/case studies/ blogs that deal with this specific problem type. Thanks a bunch.
Before clustering you first have to prepare your data.
Clustering raw data just does not work - garbage in, garbage out.
Choose appropriate features and figure out how you want to measure similarity. It's not enough to have some way to compute some number, but the similarity needs to quantify relevancy for your business, not some mathematician that doesn't care about the data.
Once you have done this homework of preparing and understanding your data, then almost any clustering can be used.