Decision Support System for Investment Portfolio Optimization Considering Individual Objectives and Risk Attitudes

Per Wong Xue

Summary

Investing has been a popular topic of interest to date. This is due to its potential for financial growth and security. Hence, having a comprehensive decision support system (DSS) remains pertinent today. However, previous research all have one flaw. They fail to address the need to incorporate the different risk measures and optimization models for investors based on their investing objectives and risk profiles.

Hence, this paper proposes an interactive DSS framework with three main features. The first feature is a forecasting system to forecast returns. Machine learning algorithms and time series analysis are used and compared against using root mean squared error (RMSE) alongside personal judgment to determine the most accurate forecasting model. The second feature is a multi-criteria decision-making (MCDM) based risk assessor. The risk assessor first uses an analytic hierarchical process (AHP) rating model to categorize the investors' risk profile. The risk assessor system then maps the risk measures available in the financial market to investors based on their profiles. The last feature is a portfolio optimizer system, where optimization models will run and output the most optimal asset allocation to the investor.

Experimental studies were conducted to map the risk measures. The effectiveness of the risk mapping and the proposed framework is then validated and tested using case studies. The results of the case studies showed that the DSS in this paper is competent in fulfilling the investors' financial needs and risk preferences. Thus, this shows the validity of the risk mapping approach, which could unlock the potential for future research for a more comprehensive DSS for investing.

However, thorough testing was missing in the experimental and case studies steps due to time constraints faced when doing the paper. This could undermine the accuracy of the findings. Therefore, future works could include more rigorous testing of the risk mapping technique. Additionally, subsequent research could focus on automating the system to determine the suitable risk measures and portfolio models to use through unsupervised learning to uncover patterns.