<p>Stock market trading is arguably the most significant topic in finance, with predicting stock prices being one of the biggest challenges because of the complex interaction between economic fundamentals, market forces and investor perceptions. The nonlinear and highly volatile nature of stock price time series data makes forecasting an even more difficult task. Although traditional methods can enhance forecasting accuracy, they increase computation complexity and prediction errors. In the present study, a novel blended scheme combining the recurrent neural network (RNN) with the grasshopper optimization algorithm (GOA) is used to address these problems. For training&#xa0;the schemes, historical data&#xa0;of&#xa0;open, high, low, close and&#xa0;volume trading&#xa0;were selected as the&#xa0;most&#xa0;significant&#xa0;features&#xa0;as these&#xa0;parameters&#xa0;are&#xa0;crucial&#xa0;for&#xa0;predicting&#xa0;the price&#xa0;movement&#xa0;of stocks&#xa0;and&#xa0;identifying&#xa0;the&#xa0;status&#xa0;of the market. Experimental results on the Nikkei 225 stock index&#xa0;data&#xa0;set&#xa0;(1 January 2013 to 2022)&#xa0;indicate&#xa0;that the&#xa0;proposed&#xa0;blended scheme provides significantly lower errors and higher accuracy in predictions. Specifically, the GOA–RNN model achieved a root mean square error (RMSE) of 119.12 and EVS of 0.9958 on the test set,&#xa0;which&#xa0;outperformed the&#xa0;traditional schemes, such as RNN and the integration of RNN with the slime mold algorithm and particle swarm optimization. These&#xa0;findings&#xa0;confirm that the&#xa0;new&#xa0;method is highly competitive and provides more accurate predictions than existing&#xa0;methods.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Estimating the changes and oscillations in the financial markets: a case study of the Nikkei 225

  • Wei Yu

摘要

Stock market trading is arguably the most significant topic in finance, with predicting stock prices being one of the biggest challenges because of the complex interaction between economic fundamentals, market forces and investor perceptions. The nonlinear and highly volatile nature of stock price time series data makes forecasting an even more difficult task. Although traditional methods can enhance forecasting accuracy, they increase computation complexity and prediction errors. In the present study, a novel blended scheme combining the recurrent neural network (RNN) with the grasshopper optimization algorithm (GOA) is used to address these problems. For training the schemes, historical data of open, high, low, close and volume trading were selected as the most significant features as these parameters are crucial for predicting the price movement of stocks and identifying the status of the market. Experimental results on the Nikkei 225 stock index data set (1 January 2013 to 2022) indicate that the proposed blended scheme provides significantly lower errors and higher accuracy in predictions. Specifically, the GOA–RNN model achieved a root mean square error (RMSE) of 119.12 and EVS of 0.9958 on the test set, which outperformed the traditional schemes, such as RNN and the integration of RNN with the slime mold algorithm and particle swarm optimization. These findings confirm that the new method is highly competitive and provides more accurate predictions than existing methods.