Neural network modelling of financial data.

• Main Author: Karolewski, M. A.
• Other Authors: Tan, Hwee Cheng
• Format: Theses and Dissertations
• Language: English
• Published: 2009
• Subjects: DRNTU::Business::Finance::Equity
• Online Access: http://hdl.handle.net/10356/20206
• Institution: Nanyang Technological University

The report deals with the application of neural network modelling techniques to two categories of financial data, namely stock price time series and financial ratios. The purpose of the inquiry is to benchmark neural network tools on financial data derived from the Singapore market, and to provide a review of the methodology and literature relating to these tools. Three specific neural network applications are considered in detail. These are (a) the modelling of stock price time series, (b) sparse modelling of financial ratios data, (c) stock price variations associated with the release of accounting information. The main conclusions which emerge from the study concern the viability of neural network techniques in the context of the Singapore market. Neural network modelling of stock price time series in isolation is apparently unable to generate useful forecasts concerning future stock price movements. However, neural networks show greater promise in those applications which involve the modelling of financial ratios, particularly in the area of data reduction. A recurring theme in the report is the difficulty of exploiting the full capabilities of neural networks in the Singapore context due to the limited availability of financial ratios data for individual industrial sectors.