We consider the maximizing travelling salesman problem (MTSP) for two
special classes of n x n matrices with non-negative entries, namely, m
atrices from M(n) and M(n, alpha) (alpha greater than or equal to 3) d
efined as follows. An n x n matrix W = [w(ij)] epsilon M(n) if w(ij) =
0 for all i,j such that /i-j/not equal 1. An n x n matrix W = [w(ij)]
epsilon M(n, alpha) if min(/i-j/=1) w(ij) greater than or equal to al
pha max(/i-j/<not equal >1) w(ij). We describe an O(n)-algorithm solvi
ng exactly the MTSP for matrices from M(n) and show that this algorith
m provides an approximate solution of the MTSP for matrices from M(n,
alpha) for alpha greater than or equal to 3 with a relative error of a
t most n/(2 alpha(n - 1)). It is proved that the MTSP is NP-hard for m
atrices from M(n, alpha) for every fixed positive alpha.