In the time series, after isolating trends and periodicity, a normalized time series is left. To check whether the data follow some well known stochastic process, model fitting is done. If the model has an autocorrelation then it is de-correlated and after de-correlation it is checked whether it behaves like white noise, or not. The article further explains how to remove auto-correlation in a time series with the help of first order autocorrelation and linear algebra framework in PCA with an example as well. Read full article at: http://www.datasciencecentral.com/profiles/blogs/how-and-why-decorrelate-time-series