Guys, I got sick yesterday and today I was still recovering, so, I’ll go slower on the “Time Series Analysis” of today, if you haven’t seen the previous post, I highly recommend since it has a custom code attached to it that shows nicely the improvements on result you can reach just by looking at smaller time frames instead of the whole series.
So today, I’ll just try to explain the next concept in a more visually appealing way, so:
Exponential Smoothing Average
Neither way, getting the mean of the whole dataset or getting the mean of a time frame are not the best choice to keep up with trends. To bring a solution to that dilemma there is the Exponential Smoothing Average (ESA). As we saw, in the Moving Average, all values were given the same weight when calculating the new value of the trend, but we obviously know that (not always) the longer an event happened, the less effect it will have in a posterior one.
So, what ESA does is decrease the value of older events exponentially when forecasting new values.
I’ll explain the formula and content of it later since I’m still really sick to provide other value today. I still recommend you run the code from the last post and try to come up with new ways to build on that.
I’ll try to bring this section at the end of my sections whenever I can, it’ll be a link to news posts that I find interesting and the first one is about VR, AR and AI applications on traveling (imagine where else they can be useful).