Формирование инвестиционного портфеля на российском рынке акций при помощи непараметрического метода - искусственных нейронных сетей

Корпоративные финансы

Artur Sarkisov¹  Elena Buyanova² 

[1] Ph.D. of the Department of Finance, HSE, Moscow, Russia;candidate of physico-mathematical Sciences, associate Professor of the Department of Finance, HSE, Moscow, Russia;

关键词: optimal portfolio;    fundamental analysis;    technical analysis;    portfolio selection;    nonparametric methods;   

DOI  :  10.17323/j.jcfr.2073-0438.11.3.2017.100-110

来源: DOAJ

【 摘 要 】

In this paper, a nonparametric method, the Artificial Neural Network (ANN), was used for analyzing 50 stocks, which are included in the calculation base of the MICEX stock index. This method allowed us to use not only macroeconomic and technical factors, but also factors with a limited data set (factors of fundamental analysis). As a result, we construct-ed an optimal portfolio with an average return of 8% higher than the market portfolio with the same risk during the period of Jan 2015 – Jan 2016.The ANN method also allows the conduction of a comparative analysis of the influence of factors on stock return. As a result, we showed that the Russian stock market has the features of a speculative market because the most important fac-tors for the stock return of Russian stocks are momentum, bid – ask spread, and the oil price. Significantly, the same fac-tors were determined in the research dedicated to the problem of constructing an optimal portfolio on the Russian stock market using the classification and regression tree (CART) method. Potential investors take into account the oil price as the key determinant of the economic environment in Russia and select stocks with high momentum and high liquidity.In this research, the ANN method was compared with another nonparametric method (CART) by solving the utility maximization problem of investors with a different coefficient value of risk aversion. As a result, the ANN method shows a strictly higher return than CART during the analyzing period.This fact could be explained by the logic of the ANN method: ANN doesn’t require numerous observations of the number of independent variables.On the other hand, the CART method requires more observations because of the minimal number of observations in one node. More variables are in the structure of CART, with more nodes in the tree, and hence, more observations are required.

【 授权许可】

Unknown