A hybrid approach for portfolio selection with higher-order moments: Empirical evidence from Shanghai Stock Exchange. (1st May 2020)

• Record Type: Journal Article
• Authors: Chen, Bilian; Zhong, Jingdong; Chen, Yuanyuan
• Is Part Of: Expert systems with applications. Volume 145(2020)
• Publication Date: 2020-05-01
• Subjects: Portfolio optimization -- Higher-order moments -- Genetic algorithm -- Machine learning algorithm; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2019.113104
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: A portfolio selection problem with higher-order moments is considered. Machine learning algorithms are applied for data analysis and prediction in the stock market. Genetic algorithm is used to solve the multi-objective optimization problem. The out-of-sample performance of our model is significantly better than those of traditional ones. Robustness is checked, compared with another two existing methods. Abstract: Skewness and kurtosis, the third and fourth order moments, are statistics to summarize the shape of a distribution function. Recent studies show that investors would take these higher-order moments into consideration to make a profitable investment decision. Unfortunately, due to the difficulties in solving the multi-objective problem with higher-order moments, the literature on portfolio selection problem with higher-order moments is few. This paper proposes a new hybrid approach to solve the portfolio selection problem with skewness and kurtosis, which includes not only the multi-objective optimization but also the data-driven asset selection and return prediction, where the techniques of two-stage clustering, radial basis function neural network and genetic algorithm are employed. With the historical data from Shanghai stock exchange, we find that the out-of-sample performance of our model with higher-order moments is significantly better than that of traditional mean-variance model and verify the robustness of our hybrid algorithm.