Learning Investor's Utility Function in Interactive Multi-Objective Investment Portfolio Optimization

Li Dongyu

Abstract

Learning has become an increasingly popular concept with a wide range of applications across various industries, but has rarely been explored in the arena of financial investment. With investors demanding a more personalised portfolio, it is critical for financial planners to have an accurate gauge of their clients’ utility functions and risk attitudes. The current industrial approach for assessing such risk attitudes is largely based on questionnaires on the clients’ demographic information, socioeconomic relationships and investment preferences. Building on existing solutions, this paper proposes a new decision-making support tool to incorporate an interactive process with the concept of learning in portfolio optimisation. In this process, the investor, also known as the decision maker (DM), could identify multiple criteria to be optimised at the same time and utilise the Multi-Criteria Decision Analysis (MCDA) tool to obtain his or her most preferred portfolio. The degree of risk aversion of the DM, often indicated by a constant A, could also be “learnt” and refined through this interactive process. As there are still research gaps in applying the concept of learning to deriving the utility function of a DM, the methodology proposed by this paper could fill up these gaps and deepen professionals’ understanding on eliciting a DM’s utility function. Moreover, this paper goes beyond simply learning the utility function, but also studies the pattern of A obtained through the interactive process to understand more about the DM’s stability of risk preference as part of the risk profile analysis.