3 Stock Market Strategy and Financial Analysis

The financial industry is a tremendous consumer of advanced computing technologies and mathematical modeling techniques, and a primary employer of computer science graduates in the New York metropolitan area.

This project focuses on creating new financial data sources and devising innovative and effective techniques for analyzing them, and providing access to our analysis for small investors over the web, so as to help them make more informed financial decisions.

Projects in financial analysis include:

• When to Buy/Sell Service - Consider the small but long-term investor who will sell a security over the next week or month, say to pay for college or make a down payment on a house. The investor can take advantage of his/her flexibility in timing to place a sell order at a higher-than-market price that they anticipate the market will reach during this period.

  But what should this price be? It is a function of market volatility and current conditions, and can be determined through mathematical analysis and the study of historical data. However, such resources have been out of the range of the small investor.

  We have developed a class of random walk models which fairly accurately models price fluctuations, and built a website to present our results (http://www.textbiz.org). This project concerns additional development of this site, with new data sources and analysis techniques applied to all major stocks/funds. Our goal is to avoid the use of sophisticated options while ensuring the small investor the fairest return on their investment.

• World Monitor Service - We are interested in aggregating, creating, and analyzing time-series data for nations, states, and corporate/financial entities from a variety of traditional and non-traditional sources. For example, time-series data such as foreign exchange rates and stock market indices provide a sense of the economic health of a given nation. Other interesting time series can be derived by aggregating other data, such as computing a daily export index for a given country as a function of its particular distribution of exports, current commodity prices, and foreign exchange rates.

  We will develop novel indices/statistics from careful analysis of web data, and study how well these indices correlate with/predict future trends. Similar measures can be computed for states in the United States and provinces in Canada, industrial sectors and individual companies/brands.

• Trend Analysis from News Streams - Another source of interesting time series data comes from analyzing news/text streams from a variety of media, and seeing how the frequencies of keywords occur, and whether in positive or negative contexts. We seek to efficiently analyze large volumes of text in order to identify new concepts and trends.

  We are currently applying natural language processing (NLP) techniques to identify interesting relationships between people, products, and companies in large news streams, and are working to expand our analysis to EDGAR financial documents and other sources.

  It would be interesting to compute a ``noise'' index for companies and other actors based on how much talk is going on about it in (a) real newspapers, (b) chat rooms, and (c) spam/press releases ("Business Wire")

There is room for several students on this project. Interested students should study the lectures from the special topics course in Computational Finance I offered in Spring 2007. Lecture notes and full course audio is available at http://www.cs.sunysb.edu/tex2html_wrap_inline$$skiena/691/.