Forecasting volatility in financial time series data with GARCH
When working with financial time series data, a common task is measuring volatility to represent uncertainty in future returns. Generally, volatility measures the spread of the probability distribution of returns and is calculated as the variance (or standard deviation) and used as a proxy for quantifying volatility or risk. In other words, it measures the dispersion of financial asset returns around an expected value. Higher volatility indicates higher risks. This helps investors, for example, understand the level of return they can expect to get and how often their returns will differ from an expected value of the return.
Most of the models we discussed previously (for example, ARIMA, SARIMA, and Prophet) focused on forecasting an observed variable based on past versions of itself. These models lack modeling changes in variance over time (heteroskedasticity).
In this recipe, you will work with a different kind...
