Abstract:
The simplex method is an iterative method for solving a linear programming problem. Its important step is exchanging an entering variable and a leaving variable, called the pivoting step, to improve the solution. Generally, the pivoting step exchanges only one entering and one leaving variable in each iteration. Therefore, if we can exchange more than one variable in each iteration, the number of iterations of the simplex method might be reduced. In this dissertation, we present a new pivot rule for the simplex method that can exchange at most three variables in each iteration, called the multi-pivot simplex method. We establish the proposed interior search algorithms to solve the special two- and three-dimensional linear programming problems to search for two or three leaving variables. In each iteration, if there are two or three negative reduced costs of nonbasic variables (for a maximization linear programming problem), then a special two- or three-dimensional linear programming problem is constructed, and the proposed interior search for two or three leaving variables is performed. By the computational results, we found that the proposed approaches can reduce the number of iterations and the running time compared with the simplex method and the double pivot simplex method based on the slope algorithm
Thammasat University. Thammasat University Library