A heuristic for finding embedded network structure in mathematical programmes

An intuitive description is given of how to recognise when a mathematical programme can be converted to a network flow problem. In practice, it is more likely that a mathematical programme with a high degree of network structure will have additional complicating constraints. Thus, a heuristic is presented for identifying a maximal subset of the constraints which can be converted to a network flow problem. Computational results show that the heuristic is highly effective, and that much larger networks are identified by seeking a general network structure, rather than limiting the search to a subset of rows giving at most one ¸+1 and one ¸-1 in each column, as has been done previously.