Like with any other classification system, there're many dimensions which you can use  to classify them. In practice some are more useful than others. For example, if you need to pick one for a given problem, most likely you'll base your decision on these 4:

• by the data model: key/value, column-oriented, document, graph (guess you're referring to these 4)
• by the query model: what kind of API it gives you
• distribution models: single-server, replication (master-slave/peer-to-peer...),  replication, sharding (can be combined with replication)
• CAP theorem: in case of partitions which one it gives you, consistency or availability (but it's more complicated than this, have a look at the paper: CAP twelve years later: How the "rules" have changed)

Article CAP Twelve years later: How the "Rules" have Changed

Dear Milan,

Thanks for such an information.

I am referring to those 4 models only but when i found these whole list of databases with various categories, i got confused!!!

so, if i say "based on data model there are 4 major categories of NoSQL databases", is it fine?

It's not clear how those 4 close the entire universe of possible data models. Where would you then put object, xml or hybrid databases? Object is sort-of like a graph but graphs don't have this notion of inheritance.

It's also not clear what do you mean by major. Are object databases a major category of NoSQL? 

That my problem is, i am getting confused when making a statement about categories of NoSQL databases. therefore, i want to know the answer from some expert.

P.S.: major, indicates main(or generalize) here...