In this paper, we propose an algorithm to decompose a geometric programming problem into a set of subproblems. Each of the subproblems is constructed with some column vectors that are chosen from the n×m exponent matrix of the given problem to satisfy each of the following :
(1) the convex hull of the selected n+1 vectors contains zero vector as an interior point;
(2) every column vector of the exponent matrix must be included in some set which consists of n+1 vectors that satisfy the condition (1).
The present paper represents a systematic choice algorithm of the column vectors. As the selection method is a pivoting method, the complexity of computation is very small.
At last some examples are dealt with to show the efficiency of this algorithm.