Comparison of Nonlinear Optimization Methods on a Multinomial Logit-Model in R

In this article we analyze several optimization methods of a real-life nonlinear optimization problem in R. The optimization problem results from fitting a multinomial logit-model. This is part of the so called Brand Simulator, a tool to forecast and simulate possible marketing mixes. In the existing optimization procedure, simple derivation-free techniques are used. The objective of this paper is to apply several promising optimization techniques in order to improve optimization with respect to time and quality. Hence a testbed was implemented in R. Results show that the proposed techniques outperform existing optimization procedures in most of the cases.