Scholarlink

The study employed a neural network for forecasting the volatility of the Nifty 50 index for the year 2025 using historical GARCH-predicted volatility data for the period of 2013 to 2024. Historical garch-predicted volatility was fed as an input to train and test four different models of neural networks with different combinations of hidden layers. The efficiency of each of the models of GARCH is tested based on AIC, whereas mean squared error was considered in evaluating the neural network model. The best model with the lowest AIC and MSE is chosen further for extending the model, the one with two hidden layers with four and three neurons whose MSE is 0.024078; this model was considered an optimal model for volatility prediction because of the lowest MSE compared to other constructed models. The forecasted daily average volatility of the Nifty 50 for 2025 is around 0.0537, which indicates moderate market fluctuations. The current study highlighted the efficiency of econometric models and neural network models in forecasting financial time series.

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

Neural Network-Based Volatility Forecasting: A Hybrid GARCH-ANN Model for Nifty 50

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jahnavi M,
  • Pooja R. Takalkar,
  • Purushottam Bung,
  • N. Nagasubba Reddy

摘要

The study employed a neural network for forecasting the volatility of the Nifty 50 index for the year 2025 using historical GARCH-predicted volatility data for the period of 2013 to 2024. Historical garch-predicted volatility was fed as an input to train and test four different models of neural networks with different combinations of hidden layers. The efficiency of each of the models of GARCH is tested based on AIC, whereas mean squared error was considered in evaluating the neural network model. The best model with the lowest AIC and MSE is chosen further for extending the model, the one with two hidden layers with four and three neurons whose MSE is 0.024078; this model was considered an optimal model for volatility prediction because of the lowest MSE compared to other constructed models. The forecasted daily average volatility of the Nifty 50 for 2025 is around 0.0537, which indicates moderate market fluctuations. The current study highlighted the efficiency of econometric models and neural network models in forecasting financial time series.