Scholarlink

Many scholars worry about the stock market, a key economic component. Many experts are exploring how to forecast stock prices and market trends. Previous prediction methods centered on geometric and neural network models, which have become admired. Although not typically utilized in financial time series, deep learning offers excellent learning abilities and is ideal for complex timeseries. Due to its cyclic structure, the LSTM network has long-term memory, making it perfect for financial timeseries anticipation. A new stock closing price anticipation framework yields better predictions than standard models in the study. Data pre and postprocessing use empirical-wavelet-transform (EWT) and outlier-resilient-extreme learning-machine (ORELM) models. Dropout strategy and particle swarm optimization (PSO) method optimize the mixed frame’s primary component, an LSTM predictor. Each hybrid framework algorithm can maximize its functionalities to improve prediction accuracy. Three hard datasets are used for predicting trials to test the model. To demonstrate the framework’s efficacy, some comparison models are chosen. Investigational results reveal that the hybrid approach proposed in the study has the best anticipation accuracy and may be used for stock market data investigation.

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

A Novel Financial Market Prediction Framework Based on Deep Learning for Accurate Trend Forecasting

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Priya Shirley Muller,
  • Yadala Sucharitha,
  • Shilpa Kottapally,
  • Gulshodakhon Ismoilova,
  • Lola Rakhimova,
  • Pundru Chandra Shaker Reddy

摘要

Many scholars worry about the stock market, a key economic component. Many experts are exploring how to forecast stock prices and market trends. Previous prediction methods centered on geometric and neural network models, which have become admired. Although not typically utilized in financial time series, deep learning offers excellent learning abilities and is ideal for complex timeseries. Due to its cyclic structure, the LSTM network has long-term memory, making it perfect for financial timeseries anticipation. A new stock closing price anticipation framework yields better predictions than standard models in the study. Data pre and postprocessing use empirical-wavelet-transform (EWT) and outlier-resilient-extreme learning-machine (ORELM) models. Dropout strategy and particle swarm optimization (PSO) method optimize the mixed frame’s primary component, an LSTM predictor. Each hybrid framework algorithm can maximize its functionalities to improve prediction accuracy. Three hard datasets are used for predicting trials to test the model. To demonstrate the framework’s efficacy, some comparison models are chosen. Investigational results reveal that the hybrid approach proposed in the study has the best anticipation accuracy and may be used for stock market data investigation.