Visually augmented high dimensional sensitivity analysis

Abstract

Route optimization models take in many parameters which results in a large solution space. Sensitivity analysis can be used to find the optimal values for these inputs. However, because each run of these models consumes a significant amount of computation resources, users must carefully choose which parameter combinations are most promising to explore through sensitivity analysis. This thesis proposes a method for visually augmented high dimensional sensitivity analysis which will allow users to make informed decisions on which parameters are tested in high dimensional sensitivity analyses. An example interface is then evaluated on the knowledge requirements for and the efficiency with which a user is able to complete tasks related to making and exploring the results of sensitivity analyses. We find that our interface is more accessible and efficient than directly accessing the database or using the database API for all but one of the tasks.