Although here I will refer specifically to C++ and Bjarne Stroustrup's naming conventions, in principle, I've seen that people use somewhat similar rules for other languages here and there.
So, the basic idea is that one should be able to distinguish standard types from user-defined types while reading the code. For instance, Bjarne Stroustrup suggests that one uses
an initial capital letter for types (e.g., Square and Graph)
which, taking into account that
The C++ language and standard library don't use capital letters
allows achieving the goal mentioned above.
But why do we need to do so? What can be the purpose of distinguishing standard and user-defined types?
I could not find any Bjarne Stroustrup's reasoning on that matter, and besides, I myself think in diametrically opposite way. :D I know, I know, "Who am I to dispute Stroustrup?" But, listen, a bunch of the C++ language features, e.g. operator overloading, serve the purpose to allow user-defined types a similar level of syntactic support as standard types. And then all this is baffled by a different naming discipline...
P.S. Not to mention that often one word is not enough to name a class and an underscore-separated word that starts with a capital letter looks so foreign.