Paper accepted for presentation at Modeling Decisions for Artificial Intelligence (MDAI) which will be held in Girona, Spain this November. It will also be one of the papers of the conference published in LNAI. My travel to this conference and a visit to GIARA is being funded by two ECR grants from Deakin University; one from the School of Information Technology, and one from the Faculty of Science and Technology. Given that fitting the Bonferroni mean cannot be fit to data using linear programming in the general case, we looked at a number of special cases.
Title: Using Linear programming for weights identification of generalized Bonferroni means in R
Authors: Beliakov, G. and James, S.
The generalized Bonferroni mean is able to capture some interaction effects between variables and model mandatory requirements. While it is a non-linear function with similar behavior to the geometric mean, it is still possible for the output to be greater than zero as long as no mandatory requirements are zero. We present a number of weights identification algorithms we have developed in the R programming language in order to model data using the generalized Bonferroni mean subject to various preferences. We then compare its accuracy when fitting the journal ranks dataset.