On 1/28/12 1:08 AM, Manfred Nowak wrote:
Andrei Alexandrescu wrote:
generally what one wants is a selection of "best of breed" stocks
that are among the top ones in a variety of categories. (Relative
importance could be assigned to each category.)
This is quite different and easier than the problem initially
stated, because ranks must not be computed.
Actually they need to be computed. I think it's possible to find cases
where rank(a, k1) + rank(a, k2) < rank(b, k1) + rank(b, k2) but alpha *
a.k1 + beta * a.k2 > alpha * b.k1 + beta.k2.
One potentially confusing issue is that importance applies to rank, not
features.
Andrei
