Scholarlink

<p>Forecasting the stock market is one of the most challenging things for investors to do to increase their profits. The objective of this study is to predict the closing price of the stock using Long-Short-Term Memory (LSTM) network modified by Sin-Cosine Algorithm (SCA), Autoregressive Integrated Moving Average (ARIMA) and Generalized Auto Regressive Conditional Heteroskedasticity (GARCH) statistical models which is called LSTM-SCA-ARIMA-GARCH model. The proposed model uniquely integrates advanced machine learning and statistical techniques to enhance stock price prediction accuracy, setting it apart from traditional models. Unlike existing methods, this hybrid model captures both linear and nonlinear patterns, effectively managing volatility and optimizing prediction accuracy through its innovative structure. This study considers data from 12 different stocks: State Bank of India Network, Oracle Corporation, Microsoft Corporation, Halliburton Company, Goldman Sachs Group Inc, Cognizant Technology Solutions Corporation, Bank of America Corp, Amazon, Badische Anilin- und Sodafabrik, Bayer AG, Deutsche Bank AG, and Saipa Automobile Company (Khodro sazi Saipa) (Khsapa). There are both daily and weekly predictions, but daily predictions are more accurate. Additionally, we utilize and evaluate four datasets from UCI (AI4I, Garment, S&amp;P, and RUSSELL) using RMSE and R<sup>2</sup> metrics, and also extended the prediction to the next 15&#xa0;days. The proposed model demonstrates substantial improvements in daily predictions, outperforming the traditional LSTM, LSTM combined with Particle Swarm Optimization, and LSTM-ARIMA-GARCH models by 83.37%, 84.05%, and 55.8%, respectively. Additionally, LSTM-SCA-ARIMA-GARCH outperforms LSTMNP, AHMPSO-LSTM, and DLWR-LSTM. This innovation holds promise for more reliable and precise stock market forecasts.</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

Stock price prediction with SCA-LSTM network and Statistical model ARIMA-GARCH

  • 期刊论文

  • 发表日期：

    2025年1月5日

查看全文
  • Homa Mehtarizadeh,
  • Najme Mansouri,
  • Behnam Mohammad Hasani Zade,
  • Mohammad Mehdi Hosseini

摘要

Forecasting the stock market is one of the most challenging things for investors to do to increase their profits. The objective of this study is to predict the closing price of the stock using Long-Short-Term Memory (LSTM) network modified by Sin-Cosine Algorithm (SCA), Autoregressive Integrated Moving Average (ARIMA) and Generalized Auto Regressive Conditional Heteroskedasticity (GARCH) statistical models which is called LSTM-SCA-ARIMA-GARCH model. The proposed model uniquely integrates advanced machine learning and statistical techniques to enhance stock price prediction accuracy, setting it apart from traditional models. Unlike existing methods, this hybrid model captures both linear and nonlinear patterns, effectively managing volatility and optimizing prediction accuracy through its innovative structure. This study considers data from 12 different stocks: State Bank of India Network, Oracle Corporation, Microsoft Corporation, Halliburton Company, Goldman Sachs Group Inc, Cognizant Technology Solutions Corporation, Bank of America Corp, Amazon, Badische Anilin- und Sodafabrik, Bayer AG, Deutsche Bank AG, and Saipa Automobile Company (Khodro sazi Saipa) (Khsapa). There are both daily and weekly predictions, but daily predictions are more accurate. Additionally, we utilize and evaluate four datasets from UCI (AI4I, Garment, S&P, and RUSSELL) using RMSE and R2 metrics, and also extended the prediction to the next 15 days. The proposed model demonstrates substantial improvements in daily predictions, outperforming the traditional LSTM, LSTM combined with Particle Swarm Optimization, and LSTM-ARIMA-GARCH models by 83.37%, 84.05%, and 55.8%, respectively. Additionally, LSTM-SCA-ARIMA-GARCH outperforms LSTMNP, AHMPSO-LSTM, and DLWR-LSTM. This innovation holds promise for more reliable and precise stock market forecasts.