Use of Neural Networks in Financial Trading and Asset Allocation
Open Access
- Author:
- Parbhakar, Manu
- Graduate Program:
- Computer Science and Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- July 10, 2008
- Committee Members:
- Yuan Xie, Thesis Advisor/Co-Advisor
- Keywords:
- Neural
Technical Indicators
ANN
AI - Abstract:
- Recent research has been focused in using neural networks for developing trading strategies and also for tactical financial asset allocation. These strategies for both tactical asset allocation and financial trading in various time horizons are based on attributes of the historic time series data – volatility (std.deviation), returns, mean etc.Further some of these trading strategies use technical indicators as inputs into the neural networks. Technical indicators are based on price action of the financial assets – like the opening, closing, high, low and volumes of any particular security. These indictors tend to capture the sentiment quotient in the financial markets and make profits from capturing that. My paper uses a number of different technical indicators to develop trading strategies and conducts comparative analysis between various neural network architectures and technical indictors to see the best fit for trading market indices and individual stocks. Further the paper uses neural networks to do tactical asset allocation. Based on the historic correlation between the assets the neural network is trained to predict when to enter or exit a particular asset class. This decision is made periodically on monthly basis. Different methodologies are evaluated to find the optimal parameter to measure the accuracy of the neural model. Finally using paper money P/L is calculated in both strategies and evaluated against a default “buy and hold” strategy