I'm curious why things like PCA/ML aren't use frequently in trading? Is there an underlying philosophy that prevent this? What I was thinking, was that if PCA worked for making money, then everyone would do it, and so it effectively wouldn't work.
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
4
ML is a very broad term. Do you mean linear regression? To you mean random forests? People use all of these approaches with various degrees of success. Bloomberg will have a story every few months about a big quant/ML fund starting or being shut down.
PCA specifically is used quite a bit in fixed income to model the underlying characteristics of fixed coupon instruments. Most of these fixed instruments can explain much of their variance from just the first 2-3 PCA components.
With equities is more challenging to use PCA as the market component is usually the largest PCA weight but then it's the wild west after that with the variance spread over many eigenvectors without a clear real-world explanation.
-
$\begingroup$ Thanks so much. Can you explain what you mean by your last sentence? What exactly is "market component" and what do you mean by its the wild west with variance spread over many eigenvectors? $\endgroup$ Commented Sep 8, 2020 at 4:21
-
$\begingroup$ Are you familiar with what pcap does? I'm going to sleep now but I'll try to find some material tomorrow. Basically pca finds common underliers that explain the movements of the underlying items. In equities there are too many factors to be useful outside of the first, which people take as general market exposure. $\endgroup$– JoshKCommented Sep 8, 2020 at 4:25
-
-
$\begingroup$ ML is a promising area in trading/investing, with a lot of interest from firms. My job does not involve ML, but I understand it is not easy to make it work. $\endgroup$– nbbo2Commented Sep 8, 2020 at 17:11