As a solution algorithm for Unbounded Knapsack Problem, the
performance analysis of density-ordered greedy heuristic, weight-ordered
greedy heuristic, value-ordered greedy heuristic, extended greedy
heuristic and total-value heuristic has been done. Empirical experiments
on different test problems have been analysed and reported. Problem
instances with a very large number of undominated items were generated
in addition to the types of instances suggested by Martello and Toth
(1990). Theoretically, the lower bound on the performance for total-value
heuristic is better than the corresponding lower bounds for the densityordered
greedy heuristic and the extended greedy heuristic as discussed
by White (1992) and Kohli and Krishnamurti (1992). The computational
tests fail to show clear superiority of any particular heuristic algorithm,
although each heuristic produces good quality solutions. If the
combination of the density-ordered greedy and the total-value greedy
heuristics are considered then the combination shows complementary
effect. A new heuristic algorithm incorporating the structural properties of
the density-ordered greedy heuristic and the total-value greedy heuristic
is developed and its complementary effect studied. It was found that the
combination of the density-ordered greedy heuristic, the extended greedy
heuristic, the total-value greedy heuristic and the new complementary
heuristic gives a better performance result than the single best heuristic in
the combination.