A Linear Algebra Correlation Stock Screener?
posted on: 2009-12-12 16:54:45 by Soo-Young
tags: site news

---

(image source)

As you may or may not know, I'm currently attending graduate school for financial mathematics. If you're unfamiliar with the subject of financial mathematics, it basically applies computer programming and high level math to trading and investment systems. This semester, the most important thing I've learned about is linear algebra and its many uses. I previously blogged about how Google uses linear algebra in its page rank algorithm, which was the first time I learned about a real world use of linear algebra.

Linear algebra can also be used to determine the correlation between any number of elements. In a lecture earlier this month, my professor was talking about how this can be used to determine how certain indexes correlate to each other. If this could be used to determine the correlation between indexes, why not stocks?

If you look at the top right corner of this site, you'll see stock quotes for the three main US indexes. These quotes come from a custom CSV file generated by Yahoo! Finance. You can define a whole range of values for these CSV files to display. I added these quotes to this site about the same time my professor gave his lecture about using linear algebra to find correlation between indexes. If I could get detailed stock information in an easy to use CSV file from Yahoo! Finance, and send that through a linear algebra correlation script, I might find some interesting stock picks.

Correlation itself doesn't seem very helpful unless you're interested in hedging your risk by buying stocks that correlate to the positions you hold (to level out your returns). This is not what I'm interested in finding out though. What I am interested in is finding stocks that have strong lagging correlation. What I mean by lagging correlation is that if stock B has strong lagging correlation to stock A, stock B will perform similarly to stock A with a certain lag period separating the two. This could potentially be used to predict the future performance of stock B based on the performance of stock A.

At this point, this is pure mental masturbation, but it might be something I'll put together in the future.