Saaty(1977) provided a table for randomly generated pairwise comparison matrices to formulate the Inconsistency Ratio. Here my confusion is: how can we find randomly $n\times n$ generated pairwise matrices?
Could anyone give me hint how I can calculate it? Thanks
Unless you have some kind of bound on the maximum size of an entry or some probability distribution on that size you can't really do this, since there's no way to make sense of "a random positive number".
But if you do have a distribution, or a bound so that you can use the uniform distribution, just choose the $n(n-1)/2$ elements above the diagonal independently, then fill in below the diagonal with reciprocals.
Possible word of warning. If you need several of these and you want them uniformly distributed among the possibilities this algorithm may fail. For example, if you choose the entries above the diagonal uniformly distributed in $[0,M]$ for some large $M$ then those entries will mostly be greater than $1$ while the entries below the diagonal will be less than $1$. That may bias any statistical conclusions you want to draw.