5

I'm trying to do a survey of stock market prediction methods, how they work and compare, for a computer science project. I know about neural networks, my project was originally going to be based on them, but after looking at the responses to this question:

Predict Stock Market Values

I figure that it would be better to look at the whole field. Can anyone show me some good resources to research?

I fear no math. Thanks.

Community
  • 1
  • 1
Mike
  • 51
  • 1
  • 2

3 Answers3

3

Gaussian Mixture Models, Dirichlet Processes, Weiner Process, Stochastic Processes, are terms that you should search at google. Neural networks will not do. However, you should take a look for no-free lunch theorems as well, both in CS and ML.

1

You appear to be interested in financial mathematics and algorithmic trading. Both of these fields are massive. Financial mathematics alone has a history over 100 years in length.

This is an incredibly open ended question, and you don't really have an indication of where to start. I suggest starting with the wiki pages on the relevant topics:

In a nutshell, financial mathematics deals with determining the expected future value of a stock or derivative via a formula. Algorithmic trading uses formulas from financial mathematics to suggest or automatically make trades.

The Black-Scholes model is widely used and is a good starting point for understanding the math behind the algorithms used in algorithmic trading. Start with it, and then move on to other models that you feel are better suited for what you want to accomplish.

Swiss
  • 5,556
  • 1
  • 28
  • 42
  • Thank you this was very helpful. I managed to find an online course in Mathematical Finance at: http://galton.uchicago.edu/~lalley/Courses/390/ Do you know any good books specifically about the algorithms, preferably with a lot of exercises in it? (Narrow down the question a little) – Mike Apr 06 '11 at 00:49
1

You should try to looking under the subject of "stock price probability" or "Swing trading techniques" or "Technical analysis" and model those. (BTW - Read about LTCM - Long Term Capital Management and see why volatility and predicting human behavior is next to impossible).

The ultimate answer is not predicting the stock price (which many believe you cannot do) but minimizing your total portfolio fluctuations and maximizing your risk reward as compared to a no-risk instrument such as a T-bill. This is under the subject of "Modern portfolio theory".

Also - look into http://www.quantmod.com/ (Start learning R)

http://www.stat.berkeley.edu/~aldous/157/Books/stock.html

Michael Thamm
  • 87
  • 1