In what context?

A maclaurin/Taylor polynomial approximates a known function, a least squares approximation is used to fit a curve to data (unless you are looking at a least squares cubic approx to a known function when the cubic will be an approximation "best" over an interval)

CB