A generalized network problem is a special class of linear programming problem whose coefficient matrix contains at most two nonzero elements per column. A generalized network problem with 0-1 flow restrictions is called an integer generalized network(IGN) problem. In this paper, we presented a branch and bound algorithm for the IGN that uses network relaxation. To improve the procedure, we develop various strategies, each of which employs different node selection criterion and/or branching variable selection criterion. We test these solution strategies and compare their efficiencies with LINDO on 70 randomly generated problems.