It is well known that the conventional subset sum problem with integers is NP-complete. What if the array elements can be any real numbers and also target sum can be any real number? Is it NP-complete still or harder (NP-hard)?
What model of real numbers are you using? If it's something like floating point, where everything is really rational, just multiply through by a common denominator and you're back in the world of integers. The bit-length is only polynomially larger than the ones you started with (the common denominator is at most the product of $n$ $n$-bit numbers, so it has at most $n^2$ bits), so nothing has changed, complexity-wise.
Beyond that, things will get more complicated – can you even tell in polynomial time what numbers are represented by the input?
However, the real-number version is definitely still NP-hard, since it has the integer version as a sub-problem.