Hill-climbing powered gated recurrent unit network to predict stock trading signals using mean average convergence divergence indicator
Journal of Theoretical and Applied Information Technology
This paper proposed a hill-climbing powered gated recurrent unit (HC-GRU) network to predict stock trading signals using the mean average convergence divergence (MACD) indicator. The proposed model was compared with the Buy/Hold Strategy, a benchmark strategy, and the traditional MACD indicator based trading strategy. All three models were evaluated in terms of the annualized rate of return, sharp ratio, and the number of profit/loss trades executed by the strategies. The experiments were conducted on twenty randomly selected stocks from the Bombay Stock Exchange, Nepal Stock Exchange, New York Stock Exchange, and the Shanghai Stock Exchange. From the experimental results, we observed that the HC-GRU outperformed the other two models in terms of all three measures. ARR obtained from HC-GRU strategy was 22.44% to 62.76% higher than the ARR obtained from the traditional MACD indicator based trading strategy and it was 14.69% to 71.72% higher than the Buy/Hold strategy. On the other side, 66.67% to 100% trades made by the HC-GRU strategy were profit trades but only 27.78% to 62.5% trades executed by the traditional MACD indicator based trading strategy were profit trades. From this observation, we concluded that the proposed HC-GRU approach is a superior strategy for automated stock trading whereas the traditional MACD indicator based strategy is not suitable for automated stock trading.
Saud, Arjun Singh; Shakya, Subarna; and Ghosh, Ashish, "Hill-climbing powered gated recurrent unit network to predict stock trading signals using mean average convergence divergence indicator" (2021). Journal Articles. 1993.