Predictive multi-period multi-objective portfolio optimization based on higher order moments: Deep learning approach

Abstract

We propose a Multi-Period Multi-Objective Portfolio Optimization model (MPMOPO). We used deep-learning approach to predict future behavior of stock returns. We consider four objectives, i.e., wealth, variance, skewness, and kurtosis and several constraints such as cardinality, budget, upper and lower limits of purchase, and diversification to address real-world situations. The investor can rebalance the portfolio through daily trade by buying or selling subject to transaction costs. We applied the proposed method in a daily closing price prediction of six stocks from FTSE 100. Goal programming method was used to solve the models. The results of statistical analysis show the applicability and efficacy of the proposed method in comparison with those methods which used historical data to form the portfolio.