Computer-assisted analysis for diagnosing infeasible or unbounded linear programs

This paper presents methods to provide computer-assisted analysis to diagnose infeasible or unbounded linear programs. This may be for retrospective analysis, after the difficulty has been detected, or for preventive analysis. The first method is by successive reduction of bounds o r primal quantities and dual prices. This is the most commonly used method in practice due to the availability of REDUCE procedures in commercial mathematical programming systems. The second method is by path or cycle generation, introduced by the author in a more general context of solution analysis. The third method is by "netforming," and the fourth is by coefficient screening. The last two methods are untried in practice, but it is suggested how they may provide a useful aid, particularly in certain computing environments. Collectively, these four methods comprise a foundation for further research