A highlight of last month was Palisade’s Summer Risk Congress in Munich, Germany. This one-day event involved case studies presented by customers, presentations of new features in @RISK5.5, and my own presentations – which concerned themselves with the use of the DecisionTools Suite for optimization modelling, and a presentation of modelling best practices in Excel and @RISK.
I have previously written several postings on best practices in Excel modelling, so won’t mention this further here. As well as highlighting some key linkages between the topics of risk analysis, decision-making under uncertainty, real options and optimization, my optimization talk tried to show that the theme of optimization runs through many of the DecisionTools Suite products, including TopRank, PrecisionTree, @RISK, Evolver and RISKOptimizer (and is not just a subject relating to the latter two products).
The talk also covered some of the challenges and common errors made when formulating optimization models in practice; these include the failure to capture the optimization trade-off within the logic of the model, that the Excel formulae required are often more complex than in more standard models (the model has to be flexible (and valid) over a wide range of input variable combinations), and incorrect optimization objectives (such as trying to optimize non-controllable inputs).
Finally, I mentioned some techniques to improve the speed of convergence of optimization models using genetic algorithms (such as those that are in Evolver and RISKOptimizer); this topic is covered in two later blog postings.
Dr. Michael Rees
Director of Training and Consulting
