Automatic Asset Selection and Allocation System with NSGA-II and Genetic Programming

Abstract

Making a good investment decision is always difficult because of the large uncertainty, randomness, and unpredictability of asset price. However, a portfolio can help investors achieve a better result with a proper allocation of good quality assets and appropriate weights. In this study, we construct an automatic asset selection and allocation system. We first apply the genetic programing to design new risk factors which can bring abnormal return based on classical factors and then use the classical factors to select stocks. After that, weight of each stock is optimized by NSGA-II with three objective functions: Sharpe ratio, skewness and kurtosis. Our factors generated through by genetic programming successfully capture the abnormal return and NSGA-II helps us maximize Sharpe ratio and minimize drawdown and shortfall. In the last, though we have achieved remarkable cumulative return based on the optimized portfolio, more efforts are needed while applying it to the real market