The term autocorrelation is used to refer to the correlation between observations in the same time series. Essentially, an autocorrelated time series is one that is correlated with a lagged version of itself. This is hugely important for the discovery of cyclical trends in a dataset. It helps us establish whether we can predict a repeated underlying pattern and add our residual variations on top of this. This can significantly help us in building our forecasts, because it provides the building block of any time series analysis -> the trend.