The goal of this project is to predict stock market volatility (as calculated by the VIX) as an indicator of economic risk using various regression techniques. As a student, this project also was a chance to learn more about time series analysis and web scraping techniques.
The business reasoning is that if you can predict market volatility you can understand risk and run your stock portfolio (or business) with the right degree of risk.
This time series analysis shows how to use many different techniques including autocorrelation and partial autocorrelation to get a good guess for ARIMA/SARIMAX parameters and using computational power to try and find the optimal parameters. I also tried Facebook Prophet to test something in addition to what we learned in class that I've heard can get good results.
At the end of the project I concluded that predicting market volatility is challenging (as expected). In picking a great challenging project though I learned a lot of techniques I could apply to more typical time series analysis like forecasting sales and other business data.
To learn more please read my presentation read the Jupyter notebooks where I cover EDA, optimizing ARIMA models, SARIMAX models and my Facebook Prophet test. I would love any comments or suggestions for things I could try or improve.