Users of simulation models rarely know which input parameters have the greatest influence on the system they are modeling. Often they do not know which combinations of parameters lead to the best and worst performance of their system.
The OptAnalysis libraries provide a suite of analysis tools to automate the discovery of influential variables, to identify and characterize multiple good and bad regions of the model tradespace, and to classify the robustness of point solutions and regions. OptAnalysis is used with simulation models or other data sets that contain a collection of input values and corresponding response values. OptAnalysis works over a diverse set of variable types including binary, integer, continuous, discrete, enumerated, and categorical variables. The software libraries provide a user-friendly API for customers who may have limited knowledge of how statistical analysis and data mining works yet who desire access to sophisticated analytic techniques.
Following an OptQuest driven search, or with a collection of model inputs and responses derived from another source, OptAnalysis analyzes the data and draws insights from it by running its suite of multivariate analysis tools. OptAnalysis draws on techniques such as regression, variable effects, regression trees, patient rule induction method, and clustering, and other data mining techniques.
The state-of-the-art analytics tools embedded in OptAnalysis are used to create easy to digest displays highlighting influential variables, good and bad regions of the model tradespace, and the robustness of point solutions and regions. Using OptAnalysis provides key insights and increases the return on investment of your simulation studies.
The OptAnalysis libraries are available for .NET for Windows and Java for all platforms.