Topic Author
Posts: 1
Joined: June 10th, 2019, 11:52 pm

Need some ideas for a thesis topic on Machine learning

June 11th, 2019, 12:04 am

I'd like to write my master's thesis in quantitative finance. Specifically, I am interested in the application of machine learning methods to problems in finance. I also enjoy coding in C++ and R in my free time, so I would enjoy to put those skills to use.

I would imagine (correct me if I'm wrong) that one way to structure such a thesis would be to simply introduce a problem X and then explore traditional methods to solving problem X, and then introducing some novel machine learning methods to solve the same problem, and then apply and compare the various methods under various settings and present the numeric results.

My problem is, I don't have a clue what the problem "X" could be, and what novel new machine learning ideas have been introduced that I could talk about.

So I would appreciate it if I could be given some ideas or pushed in the right direction.

My current strategy is just reading some papers on Arxiv under "Computational Finance", right now I'm trying to digest this one: (Machine Learning Tree and Exact Integration for Pricing American Options in High Dimensions)
User avatar
Posts: 10205
Joined: December 19th, 2001, 4:01 am
Location: California

Re: Need some ideas for a thesis topic on Machine learning

June 11th, 2019, 6:14 pm

Just thinking out loud, here's an idea. The official arbiter of US recession/expansion periods is the NBER. As you can see from here
Q: Typically, how long after the beginning of a recession does the BCDC declare that a recession has started? After the end of the recession?

A: The committee's determination of the peak date in December 2007 occurred 11 months after that date and the committee's action in determining the trough date of June 2009 occurred 15 months after that date. Earlier determinations took between 6 and 21 months. There is no fixed timing rule. The committee waits long enough so that the existence of a peak or trough is not in doubt, and until it can assign an accurate peak or trough date.

this official determination takes place with a significant lag. So, the ML challenge is *not* to predict when a recession might start (probably too hard), but reliably declaring that we are currently *in* a recession once it's already started, but not yet dated by the NBER. Of course, there will be traditional regression and other statistical approaches to this problem also. I'm sure some googling should turn up the latter.

Beyond that, there are lots of financial things that can be found googling for Machine Learning and _____ : various fraud detections, bankruptcy prediction, etc. Your main problem, as a master's student, is likely figuring out what the 'state of the art' for the traditional approaches really is. Use your advisor for direction on that.

Finally, one piece of advice: don't pick a topic where you are trying to use ML to predict something well-known to be essentially unpredictable, like short-term stock returns. 
User avatar
Posts: 62381
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam

Re: Need some ideas for a thesis topic on Machine learning

June 13th, 2019, 7:21 pm ... ure_Review

Maybe credit risk with SVM, Naive Bayes etc.

And I would use C++ libraries, not Python at this stage, e.g. OpenCV.
User avatar
Posts: 2313
Joined: September 30th, 2015, 8:30 pm

Re: Need some ideas for a thesis topic on Machine learning

June 15th, 2019, 10:47 pm

Sentiment detection using ML for natural language processing, if you're feeling brave.

PW by JB has been "Serving the Quantitative Finance Community" since 2001. Continued...

Twitter LinkedIn Instagram



Looking for a quant job, risk, algo trading,...? Browse jobs here...