In the previous article, we extracted a set of nine flow patterns from a two-year dataset (2018-2019). However, now there is no direct mapping between the day of the week and patterns. The rules for assigning a pattern to each day are more complicated. Obviously, if we have the measured flow of each day, we can just calculate the distance between each day and each pattern and there we have it. However, the true value lies in doing it in advance, without the measured flow. We can use all the work that we saved in using clustering to extract patterns when we need to predict how to assign a pattern.