That's a remnant from the days where the FPU was on a physically separate chip. This chip was numbered XXX+1 where XXX was the CPU model. So the math coprocessor for a 80286 was called an 80287, the 80386 had an 80387. The 80486-DX (but not the -SX) was the first model (if I remember properly) to have an integrated FPU. If you went cheap and got the -SX, you could physically install a math coprocessor next to the main CPU at a later date.
These days I'd call the floating point unit just "FPU."