Ideas for MSc dissertation...

[AdrenoCapital]

I'm currently undertaking a conversion masters in computing and have to start thinking about what I'll be doing for my dissertation. Background wise, I have a BSc in Mathematics and the masters has given me one semester of Java programming skills. The course also includes XML, database design and SQL, so using these are an option as well. I trade US equities part-time and have been for the past 3 years with some success (mostly technical analysis and tight money management).

Ideally, I'd like to combine my interests in maths, programming and trading when coming up with a project idea. As my programming knowledge is quite basic, I doubt I'll have the skills necessary to pull off anything cutting edge, but would still like to stretch myself with a demanding topic. The trick will be coming up with something that I find interesting without being overly ambitious to complete in the 3-4 months I have to do it in. I'll also probably be undertaking the C++ for FE cert too over the next few months, so using C++ instead of Java is a possibility.

As a long time reader of this forum, I'd really appreciate your input and I look forward to reading what you think might be an option for me.

 

[Ken Abbott]

Take 8 portfolios: equity-heavy, FI-heavy, commodities-heavy, FX-heavy, all with and without optionality. Run them through V/CV VaR, HistSim VaR, and MC VaR using 1, 2, and 4 years of data for the VaR. Compare results at 99%, 95%, and 90%. Try using weighting algorithms (DWAE, exponential) to see how your results change. Test the results against some actual portfolios. Test VaR / backtests pre- and post-crash.

 

[aresiusXXIII]

I also looking up for some ideas. Next month i will start last semester in 2 bologna cycle. I want to do something from financial engineering but I must include evolutionary algorithms (GA or GP) or neural networks in my work.

 

[Ken Abbott]

Alternatively, do PCA, pre and post crisis, on swap curves for usd, jpy, eur, cad, chf, gbp, sgd, hkd, aud, nzd and any others of interest. What is hedge efficiency like using the 1st PC? 1st and 2nd PC? Compare with sovereign YCs and perhaps to oil curves.

 

[AdrenoCapital]

Thanks for your input Ken, definitely interesting projects. I'd love to do a PCA and get stuck into some decent stats work, but the dissertation I've got to pull together has to be heavier on the coding side than the mathematical analysis side. I don't think they're expecting me to come up with anything earth shatteringly new either, just something complex enough to show off the knowledge I've acquired and a bit more research on top.

I was thinking of maybe coding an trading system on the following lines:

1. getting the raw data and putting it into a database table
2. analysing the data using some technical indicators
3. using a neural network in combination with some belief functions to produce buy and sell signals
4. modelling the executions
5. monitoring trade performance

I think I could build something that would do the steps above, albeit in a very simplistic form. Most of the effort and complexity I'd put into step 3 as the rest is pretty straight forward.

Do anyone have any pointers, based on their own experience of doing something similar as to how realistic this would be and any lessons learned from previous implementations that would be helpful for me?