Scaling an Optimization Engine for a Decision Support System

The objective of this survey has been to determine whether an existing linearprogramming model could be used to implement a system for the optimizationwith regards to the matching of transports and requests. Realdata from Härnösands harbour was enhanced and used to simulate scenariosthat would be used and applied within the system. The LINGOsolverwas integrated within a system that was developed in .Net to addressissues around data in- and output, data persistence, and stability.Performance tests revealed that the system and model performed inadequatelyon larger datasets. A simpler model was developed that handlea subset of the types of datasets handled by the first model, a subsetwhich includes the data collected from Härnösand harbour. The simplermodel was implemented with the AMPL/CPLEX-solver software, and furthertests showed that this provided a much improved performance to thesystem. These results indicate that further investigation might be requiredregarding the differences in performance between combinations of differentsolvers and models, and that a complete and production ready systemis likely to benefit from incorporation of various models that can be appliedto various characteristics of the input data