SYMBOLIC AND GRAPHIC CLASSIFICATION IN INTERACTIVE MULTIOBJECTIVE OPTIMIZATION SYSTEM WWW-NIMBUS
Katja Kaario, Kaisa Miettinen, Marko M. Mäkelä, Pertti O. Saariluoma · 2004
In interactive multiobjective optimization systems, the classiflcation of objec- tive functions is a convenient way to search for the best possible compromise solution. In classiflcation, the user specifles what kind of changes of objective function values (in relation to the current values) would be desirable. However, it is not clear how the classi- flcation should be implemented to serve the user in the best possible way. The uppermost goal of developing the usability of such systems is to support users during the classiflcation process as well as possible. We wanted to study the interactive multiobjective optimization system WWW-NIMBUS and especially the usability of graphic and symbolic displays, its classifying options. In particular, we wanted to clarify the difierences between the options. Furthermore, we wanted to gather information about the di-culties that users might have. We carried out experiments where subjects solved multiobjective optimization problems with WWW-NIMBUS using both the symbolic and the graphic displays to classify the objective functions. Thus, we obtained important information about how people use the classiflcation options and how the options could be improved. Here we summarize our flndings and also present some suggestions for developing WWW-NIMBUS further.