[4Rin1] Interactive 2

[4Rin1-20] Deep Portfolio Optimization of Residual Factors in the Stock Market

〇Kentaro Imajo¹, Kentaro Minami¹, Katsuya Ito¹, Kei Nakagawa² (1.Preferred Networks, Inc., 2.Nomura Asset Management)

Keywords:Finance, Time series, Deep learning

Developing a profitable trading strategy is a central problem in the financial industry. In this paper, we develop a technique for portfolio construction based on residual returns, which is a financial quantity obtained by hedging out common market factors in stock market. Our proposed method extracts information about residual factors by a simple spectral decomposition method. We also propose a novel neural network architecture that reflects well-known scale invariance structures of financial time series. We show the effectiveness of our proposed method by numerical experiments on Japanese stock market data.