Machine Learning in Finance

[Jack1991]

I thought the book was better than a lot of stuff out there, although I did find it rather odd how he only used classification and completely ignored regression. Personally I have found there is a lot of information loss when using classification especially at non-HFT time scales. Maybe the info in the book is more useful with the assets he trades but it's definitely missing some necessary steps if trying to do ML in equities which is what I trade.

So the large body of research regarding financial machine learning is related to classification rather than regression. The typical regression models used are CAPM and APT related, and a large amount of those use linear models, very dated. The financial ML papers out there focused on regression are pretty sparse.

Re Kevin's comment: de Prado just won Quant of the year 2019, is the leading researcher in financial ML and if anything he probably has a Nobel Prize waiting for him in the future. He is the head of ML at AQR, has made millions of dollars for Guggenheim Partners, did his Postdoctoral research at Harvard. I don't think a more able and credible person exists. Kevin is of course entitled to his own opinion.

 

[Kevin Schmit]

[Lopes de Prado] probably has a Nobel Prize waiting for him in the future...

When the acolytes start pre-announcing the Nobel, its a near sure sign.

Care to wager on that Jack? 10 year window on the bet. Just out of curiosity, what thesis has he articulated that is even remotely Nobel-worthy?

I'm still waiting on the Nobel Committee (Riksbank) to recognise ProfLogic.

As for regression vs. classification, I won't comment other than to observe the ML encompasses both, not all regression is linear OLS (and not all OLS is "linear"), and there are many ML-in-finance papers that use regression models (it depends on your deinfition of "sparse," I suppose).

 

[ph1l]

I thought the book was better than a lot of stuff out there, although I did find it rather odd how he only used classification and completely ignored regression. Personally I have found there is a lot of information loss when using classification especially at non-HFT time scales. Maybe the info in the book is more useful with the assets he trades but it's definitely missing some necessary steps if trying to do ML in equities which is what I trade.

In my experience with machine learning for financial forecasting, classification (using genetic programming, recurrent neural networks, or k-nearest neighbor) was always more accurate for training and future predictions than regression (using genetic programming or recurrent neural networks).

Maybe researchers have similar conclusions, so they focus on improving the methods that already have better outcomes.

 

[sle]

He is the head of ML at AQR, has made millions of dollars for Guggenheim Partners, did his Postdoctoral research at Harvard.

I was curious today, so I reached out to someone involved in the field. Apparently, people at Guggenheim have a different opinion of his work there. The operative words were "useless" and "waste of time".

I can't comment directly, not being involved in anything ML-related. My only thought would be that he worked for an asset management firm and is now working for another one. That certainly means that he is a good salesman (of himself and the AM services), but does not necessarily have any alpha to speak of.

 

[Kevin Schmit]

I reached out to ... Guggenheim ....operative words were "useless" and "waste of time".

Ditto Tudor.

And when you get an answer that clear it usually means the subject of enquiry is a complete fool. The usual answer is something like "he's no longer here, that should tell you how much we valued his work," or just a shrug and a broad grin which tells you the same thing.

Note also the way that Jack credits de Prado with making "hundreds of $ millions" for Guggenheim without citing any actual evidence of that. This is common with cranks' disciples. Reminiscent of how AsiaProp (Grul) used to claim the same thing about his option guru Taleb at Paloma, when the truth was decidely otherwise.

Was de Prado's book published by Wiley? Let me check... Yes, it was. Of cource it was!

Edit: Jack has more likes than posts. He's the new Xela.

 

[tommcginnis]

Was de Prado's book published by Wiley? Let me check... Yes, it was. Of cource it was!

((I'll admit -- back a million years ago, Wiley was one of my favorite publishers. But about 10 years ago, I started looking at some of their titles/authors, and thinking, "Huh?? *That* one slipped through the cracks..." Your *very* funny comment..... well, it kinda cements what I've not been willing to think: that they're schlockmeisters. That's fucking sad. I'm going to try and keep an open mind, but your comment certainly skews my starting point. Damn.))

 

[Jack1991]

Can anyone explain to me why its so important to trash someones name on a forum? Is this the kind of culture you want to encourage? We did some good work based on his research, showed that it worked out of sample. We are excited and sharing it with the community and instead of opening a dialog and sharing ideas we are throwing stones?

I wont have any part of this. I am sorry I ever came here. I hope that the moderators remove this post.

[Edited by Magna:] At OP's request this thread has been closed.