Autocorrelation and White Noise - Applied Time Series Analysis in Python and TensorFlow

👉 Get the course at 87% off: https://www.udemy.com/course/applied-time-series-analysis-in-python/?couponCode=TSPYTHON2021 Email me for a coupon if the one above expired: peixmarco@gmail.com ------------------------------------------------- Before we dive into predictive modelling, we need to cover the concepts of autocorrelation and white noise, as they will come up often later on. We know that correlation measures the extent of a linear relationship between two variables. Therefore, autocorrelation measures the linear relationship between lagged values of at time series. This means that we have multiple autocorrelation coefficients, each corresponding to a different lag. We usually plot the autocorrelation function or ACF. It is a scatter plot, with the lag on the x-axis and the autocorrelation coefficients on the y-axis. We also plot an interval of significance to help us determine if an autocorrelation coefficient is significant or not. Note that when there is a trend, the autocorrelation will be high for small lags, and gradually decreases as the lag increases. Therefore, we can see here that our ACF plot clearly shows a time series with some kind of trend. Note also that the autocorrelation will always be 1 at lag 0. Now, if a time series has no autocorrelation, then it is called white noise. White noise is a purely stochastic or random process. The ACF plot should not show any significant autocorrelation coefficients.