Computer Science EE on Forex and applicable AI techniques.

[ac1998]

Hi,

 

For my Computer Science Extended Essay I had drafted the following research statements:

 

1) Comparing the speed and accuracy of Artificial Neural Networks and Machine Learning Techniques in forecasting daily USD/INR and USD/GBP rates.

 

2) Investigating NP-completeness of Hexadoku by comparing algorithmic efficiency for generating solutions.

 

My teacher, instead of providing his own concrete feedback, put these topics on the OCC. The general comment that these topics received on the OCC are as follows:

 

1) Forecasting is really vague as mentioned previously. It is really important to keep in mind the timeframe particularly of such a volatile market as Forex. Forex Market operates 24 hours a day except weekends and the exchange rate depending on the currency pair may fluctuate as often as every second. Bank’s currency exchange rates fluctuates less frequently, but might still change several times during bank’s operating hours. So it is not clear what the “daily rate†is.

 

2) â€œComparison of the efficiency of algorithms†related topics are definitely more suitable for Computer Science, which is also indicated in the EE Guide. However it is important to choose a topic which is â€worthy of investigation†(Criterion B). What is the “significance†of solving Sudoku or its variations?

 

Can someone help me rectify either of these questions to meet the suggestions. Thank you very much.

[Jambola2]

1- Decide what you want.

Daily rate could refer to opening and closing rate, which is usually a fixed value. So, it could become "Comparing the speed and accuracy of Artificial Neural Networks and Machine Learning Techniques in forecasting daily USD/INR and USD/GBP closing rates."

2- I've been looking into NP-completeness for my EE too! You can use the P vs NP problem as justification for significance of proving sudoku is NP-complete.
In short, the P vs NP problem asks "Can any NP-complete problem, which can be verified to be correct in a polynomial time be also solved in polynomial time?"
Do some research on the P vs NP problem and what will happen if it is proven that P = NP.

 

Investigating if Sudoku is NP-Complete or not can be helpful. If it is NP-complete, and an algorithm can be created which solves sudoku in polynomial time, then it can be expanded to related NP-complete problems, and could possibly be used to solve the P vs NP problem.

 

In short, if sudoku is NP-complete, it could possibly solve one of the millennium math problems, and have huge implications.

 

I don't see how to change the second question, just talk to your teacher about the relation between NP-completeness and the P vs NP problem.

 

[ac1998]

Thanks a lot for your reply Jambola2, your feedback has been the most constructive since I started a hunt for my research question/statement. However, sudoku has already been defined as NP complete and can be solved quite quickly using parameterisation and randomisation. 

[Jambola2]

Thanks a lot for your reply Jambola2, your feedback has been the most constructive since I started a hunt for my research question/statement. However, sudoku has already been defined as NP complete and can be solved quite quickly using parameterisation and randomisation. 

 

Yup, meant hexadoku then.

Anything which is proven to be NP-complete can be hypothetically used towards the P vs NP problem.

[ac1998]

It has already been generalised that any of sudoku's variant of the form n^2 * n^2 can be solved in polynomial time. I doubt I can go ahead with that topic.

[ac1998]

On the other hand, there are no formal algorithms explicitly defined(apart from a couple of hexadoku solvers online).