You shouldn't expect others to do your homework, but help is surely available.
First step would be a clear specification. You can design a divider to give fractional result bits (bits right of the decimal point), however this has nothing to do with floating point number representation. Floating point involves an exponent, showing how many bit positions the result mantissa has been shifted. You should also clarify if signed or unsigned division is intended.
A simple way to get 8 integer and 4 fractional result bits would be to extend the dividend by 4 LSB zero bits and then perform a regular integer division.