Final Report - Peer Review (kc594)

[kc594]

The goal of this project is to take the CBOE Volatility Index and other related information surrounding the VIX to first be able to predict the movements and fluctuations of the VIX, and second to use these to create a trading strategy on VIX ETF.

Some highlights of the project:

1. I am not any sort of expert in the financial field, but their introduction and reasoning of the features they were including were very well explained and made sense even to someone unfamiliar with the area.
2. Along with this, the report was well organized and clearly broken up into sections and subsections and labeled distinctly to make it easy to follow and see the progression of steps you took to complete the project.
3. Very thorough in their methods to find additional information to use as features, from many different sources. Once they had information from these sources, they methods they described to clean and aggregate the information into features also seemed extremely thorough, making their project and results to come seem trustworthy to a reader.
4. Used many feature engineering techniques, apart from ones we learned in class, giving me confidence their methods were well thought out for this project to give the best possible results. Also tried many different loss functions from class as various models to compare.

Some room for future improvements:

1. Graphs and visuals were well annotated (such as figure 1 with descriptions of the events occurring at specific times), but including axis labels or more detailed captions describing what values are being graphed would have been helpful to a reader looking to be convinced of claims in the report.
2. It is talked about that some features have similar distributions as can be seen in box plot, although it is not clear to the reader where these are in the graph as the writing is too small. I am also concerned that although they may be correlated, the combination of the two may have predictive power for the result as we have discussed in class, and dropping one would lose this. Did you try running with and without removing the seemingly correlated features to see if it happened to improve accuracy?
3. The report very clearly spelled out the data cleaning and model building for the part of trying to predict the VIX movements, but when you jump to the second part of your proposal, creating a trading strategy, it is much more brief and less clear about what you are using and why you are choosing what you do (maybe this is due to my lack of knowledge regarding VIX, but it seems much more rushed at the end after a very detailed report).

Overall, this is a very impressive report and very well done.