Where is the name "coset" in group theory from?

The answer is relatively simple. It does appear that G.A. Miller did indeed originate the use of the term in his early publications on group theory. It literally means: "co-set". The prefix co- is from the Latin "com-" meaning (among other things) "together with" (as an example, the Spanish derivative "con" simply means "with"). A "co-set" of a subgroup H of a group G, is a subset of G "occuring with" H, and sharing the most important property of H as a set, which is its cardinality. It is perfectly natural to think of the cosets gH of a subgroup H as "sister sets" of the subgroup H, each one formed by multiplying H by some element g of the parent group G. This meaning of co- is the same one we use in such words as co-pilot, or co-worker.

I think it doubtful that "co-" was intended as an abbreviation for "complementary" or "common", as the modern penchant for abbreviation was not so common at the start of the 20th century.

In any case, in Theory and Applications of Finite Groups (1916), Miller indicates that he originated the term, the previous one being Nebengruppen, which I believe translates roughly as "subgroup". Perhaps a better term might have been "translate" of H, a term used in the study of topological groups (esp. Lie groups). It appears that the original use of the word "co-set" was its current use.

The utility of the concept of cosets extends far beyond Lagrange's Theorem. For (a rather simple) example, in the integers, the coset $1+2\mathbb{Z}$ is the set of all odd integers, which is often very useful as being considered as "a single entity".

Gallian's "Contemporary Abstract Algebra" says that Galois invented the concept of a coset in 1830, but the name coset was not used until 1910 by G.A. Miller.

See Mathword http://jeff560.tripod.com/mathword.html for first use of this.