You are here

Comparitive Study of Supervised and Unsupervised Learning Methods in Forecasting the U.S. 30-Year Treasury Bond Yield

Title: A Comparitive Study of Supervised and Unsupervised Learning Methods in Forecasting the U.S. 30-Year Treasury Bond Yield.
Name(s): Powell, Nicole Andrea, author
Foo, Simon Y., professor directing thesis
Meyer-Baese, Anke, committee member
Weatherspoon, Mark H., committee member
Department of Electrical and Computer Engineering, degree granting department
Florida State University, degree granting institution
Type of Resource: text
Genre: Text
Issuance: monographic
Date Issued: 2008
Publisher: Florida State University
Place of Publication: Tallahassee, Florida
Physical Form: computer
online resource
Extent: 1 online resource
Language(s): English
Abstract/Description: The prediction of any aspect of the future has always fascinated mankind because of the possible benefits of this knowledge, especially financial benefits. From year to year, many stockholders would like to be able to know if the price of their commodity will increase or decrease, and in turn this prediction may help them in the decision to buy more or sell what they currently have. It is widely known that the stock market is a volatile and complex entity which is affected by various factors such as government policies, political situations, public events, internal company politics and much more. However there is no way of knowing exactly which factor will affect stock prices and how much the price will be affected. Financial trend forecasting is a major component in corporate finance because predictions of future prices, indices, volumes and several other values are often incorporated into the economic decision-making process for a particular company. For the average investor financial trend forecasting would mean a greater profit (or smaller loss). To recognize specific trends for forecasting capabilities, it is important to develop a method for eliminating speculation and to investigate new algorithms for detecting patterns. Although there are many different approaches available, in this thesis a comparison between an unsupervised classification technique, namely K-means clustering, and supervised learning algorithms, namely support vector machines and radial basis functions, will be performed. The three pattern recognition systems will be tested against real-world data concerning the U.S. 30-Year Treasury bond yield. Determining the yield trend is approached as a technical analysis problem for this particular study: ignoring underlying factors and focusing on finding patterns directly from historical data. The results from each of the three networks are compared and analyzed. The performance measures analyzed include accuracy percentages, return on investment ratios, as well as capital gains/losses. From this analysis, a general network model can be decided upon to forecast the U.S. 30-Year Treasury bond yield.
Identifier: FSU_migr_etd-0447 (IID)
Submitted Note: A Thesis submitted to the Department of Electrical and Computer Engineering in partial fulfillment of the requirements for the degree of Master of Science.
Degree Awarded: Fall Semester, 2008.
Date of Defense: October 24, 2008.
Keywords: Financial Forecasting
Bibliography Note: Includes bibliographical references.
Advisory Committee: Simon Y. Foo, Professor Directing Thesis; Anke Meyer-Baese, Committee Member; Mark H. Weatherspoon, Committee Member.
Subject(s): Electrical engineering
Computer engineering
Persistent Link to This Record:
Owner Institution: FSU

Choose the citation style.
Powell, N. A. (2008). A Comparitive Study of Supervised and Unsupervised Learning Methods in Forecasting the U.S. 30-Year Treasury Bond Yield. Retrieved from