I worked on a project regarding statistical properties of crude oil / natural gas market/ cointegration modeling in presence of structura l breaks in time series. I even performed causality analysis post crisis
Practically, cointegration existed prior to 2009 until a decouling effect took over due to introduction of hydraulic fracking and horizontal drilling. Increased supply of shale nat gas depressed these prices considerably.
If you slide all the way down, you will notice i performed regime switching in presence of co-integration and found should regime shifts underly changes in the time series.
How about recommending me based on practical work/ research experience ?
You're not going to find a good pair for crude. Use pairstradinglab.com to search for pairs based on cointegration, R squared, etc. Given searching for pairs based on a certain stat isn't extremely relevant in industry, for a thesis, you can find a few good ones.
With that said, do some work on using neural nets to predict share price/value (whatever angle you wish to take) based on earnings reports. Hint: use R, there is already a couple of good nnet packages available. From my expirence, academics like the use of neural nets if you can back up your model.