Please use this identifier to cite or link to this item:
Title: Deep RBM neural network for financial forecasting
Authors: Lyu, Yuetong
Keywords: Engineering::Electrical and electronic engineering
Issue Date: 2021
Publisher: Nanyang Technological University
Source: Lyu, Y. (2021). Deep RBM neural network for financial forecasting. Master's thesis, Nanyang Technological University, Singapore.
Abstract: Stock forecasting is complicated with its non-linearity and high noise, and current forecasting models are mainly based on time-domain analysis. Wavelet analysis is a time-frequency domain analysis method that is widely used in signal processing. DBN (Deep Belief Network) is a type of neural network made up of RBM (Restricted Boltzmann Machine) stacks. This dissertation proposes a wavelet-DBN model that can be used for financial forecasting. Firstly we regard stock fluctuations as a non-stationary signal and perform wavelet decomposition (db5 in this dissertation). Then we separate the low-frequency components that can characterize the primary trend of the original signal and the high-frequency components that can characterize high-frequency changes, choose a part of the decomposed components as the input of DBN to perform the prediction. Thus, extract the components that carry more important information and discard ”noise” components. The experiment result shows that such extraction can improve the accuracy of the DBN model during forecasting. When using the low-frequency components only and discarding all the high-frequency components during the forecasting of the Shanghai Composite Index, the error rate can be limited to lower than 0.8%. Finally, this dissertation discusses and explains the influence of the choice of input components on the model accuracy on the Shanghai Composite Index and GEM Index.
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Theses

Files in This Item:
File Description SizeFormat 
Dissertation_Lyu Yuetong.pdf
  Restricted Access
1.53 MBAdobe PDFView/Open

Page view(s)

Updated on Jan 21, 2022

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.