Autocorrelation represents the degree of similarity of a time series and a lagged version of itself over successive time intervals. It is a very important concept as it measures the relationship between a current value and a corresponding past value. Thus, it has many valuable applications in time series forecasting; for example, to match trends and relationships in prices, stocks, returns, and so on.
We want to use autocorrelation to determine if we can reliably identify causality and trend – or if, on the contrary, we're dealing with a random walk model. A random walk would imply that the values in the time series are randomly defined, and this would imply that there's no relationship between past and present values. The ...
