Scholarlink

<p>Long short-term memory networks (LSTM) are the most often utilized algorithm for predicting time series. To improve efficiency, LSTM networks often add layering to acquire high-level information. But the network gets worse and harder to train as the depth grows. Due to the accumulation of errors between layers, and the vanishing of gradients, issues may arise. A stacked residual LSTM model for predicting the price of options is proposed in this article as a solution to this issue. This research utilizes a novel deep learning approach to efficiently and precisely calculate the prices of options. Through the residual connection, multilayer gradients and propagating errors can be centered. Consequently, it simplifies the optimization process of the deeper system. This method allows for the effective stacking of additional layers of LSTM, resulting in improved prediction precision compared to traditional approaches. The research uses both Hypothetical data and Real market data of European call options with NIFTY option contracts as its underlying asset. Additionally, by introducing noise into the data, the model’s robustness is evaluated. Our approach outperforms popular deep learning algorithms like artificial neural networks (ANNs) and long short-term memory (LSTM) in terms of prediction precision, and extensive computational tests highlight the deep ResLSTM model’s superior accuracy in option pricing.</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

Leveraging stacked residual LSTM for enhanced option pricing accuracy

  • 期刊论文

  • 发表日期：

    2025年10月1日

查看全文
  • Akanksha Sharma,
  • Chandan Kumar Verma

摘要

Long short-term memory networks (LSTM) are the most often utilized algorithm for predicting time series. To improve efficiency, LSTM networks often add layering to acquire high-level information. But the network gets worse and harder to train as the depth grows. Due to the accumulation of errors between layers, and the vanishing of gradients, issues may arise. A stacked residual LSTM model for predicting the price of options is proposed in this article as a solution to this issue. This research utilizes a novel deep learning approach to efficiently and precisely calculate the prices of options. Through the residual connection, multilayer gradients and propagating errors can be centered. Consequently, it simplifies the optimization process of the deeper system. This method allows for the effective stacking of additional layers of LSTM, resulting in improved prediction precision compared to traditional approaches. The research uses both Hypothetical data and Real market data of European call options with NIFTY option contracts as its underlying asset. Additionally, by introducing noise into the data, the model’s robustness is evaluated. Our approach outperforms popular deep learning algorithms like artificial neural networks (ANNs) and long short-term memory (LSTM) in terms of prediction precision, and extensive computational tests highlight the deep ResLSTM model’s superior accuracy in option pricing.