Sensitivity validation of a fuzzy system for asset allocation

Abstract

This paper shows how a fuzzy model for asset allocation can be validated. For this purpose, special instruments are developed that allow such testing for models developed for advisory and diagnostic tasks. The procedures and instruments are presented using the example of a fuzzy logic-based model, which is hierarchically structured and which aggregates in a rule-based manner. It was developed as a pilot model of a “Robo-advisor” with the professional support of a major bank which is globally active in asset management. The validation examples presented show that the instruments can be used to comprehensively identify the extent to which the diagnoses and recommendations of a fuzzy model correspond to those of the expert whose decision-making behavior was depicted in the model. In addition, it can also be determined whether the proposals of the model change in the same way as the expert changes his proposals when the parameters of an investment project change. The validation results prove that fuzzy concepts enable the development of a decision support model that can complement the investment advice of financial institutions in a valuable way.