Security Index Forecasting Using Multi Step Ahead Wavelet Neural Network and Wavelet Garch : A Comparative Study
Published: 2009
Author(s) Name: Swarnava Mitra, Atanu Das
Locked
Subscribed
Available for All
Abstract
Inclusion of market noise is a common assumption in financial time series analysis. Removal of such noise significantly improves the forecasting performance of a model. This paper has used wavelet transform technique to denoise a financial time series. After denoising a multi step ahead forecasting technique based on a feed forward neural network is used for forecasting on 5 days,
10 days and 15 days horizon. The performance of the wavelet neural network model is then compared with wavelet GARCH(1,1) model empirically with data from Indian security market. The RMSE results show that former outperforms the later.
keywords : wavelet transform, feed forward neural network, early stopping, GARCH.
View PDF