The tolerance approach to sensitivity analysis in network linear programming

From MaRDI portal





The network linear programming problem is defined as: \[ \min imize\quad \sum^{n}_{j=1}\hat c_ jx_ j\quad subject\quad to\quad \sum^{n}_{j=1}a_{ij}x_ j=\hat b_ i,\quad i=1,...,m, \] where \(A=(a_{ij})\) is the node-arc incidence matrix of a network, \(\hat b_ i\) is the demand or supply of node i in the network \(\sum^{m}_{i=1}\hat b_ i=0\). In the tolerance approach to sensitivity analysis the problem \[ \min imize\quad \sum^{n}_{j=1}(\hat c_ j+\gamma_ jc_ j)x_ j\quad subject\quad to\quad \sum_{j=1}a_{ij}x_ j=\hat b_ i+\lambda_ ib_ i,\quad i=1,...,m, \] is studied where \(\gamma_ j\), \(j=1,...,n\) and \(\lambda_ i\), \(i=1,...,m\) are parameters. The problem is to determine ranges for \(\gamma_ j\) and \(\lambda_ i\) such that the optimal solution remains optimal whenever the parameters are simultaneously varied within their corresponding ranges. The maximum absolute value of each parameter is called the maximum allowable tolerance. It is shown that this value can be efficiently calculated for the two cases considered in the paper, namely the tolerance approach for the objective function only or the tolerance approach for the right-hand side only.











This page was built for publication: The tolerance approach to sensitivity analysis in network linear programming

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3799833)