There are a lot of aspects to time series analysis. One method is called Holt-Winters and uses exponential smoothing to handle trend and seasonality. If you want, you can read about it in the original paper called “Forecasting Sales by Exponentially Weighted Moving Averages” by Peter R. Winters in Management Science 1960.

The method has been included in some analytics platforms. If you are using R, you might want to check out https://towardsdatascience.com/time-series-forecasting-in-r-with-holt-winters-16ef9ebdb6c0 and the code at https://github.com/bamattis/Blog/tree/main/R/TimeSeries

However, this challenge will use python, and to avoid the trouble of installing python (if you haven’t already done so) this is a great opportunity to try Google Colab.

First, read through the excellent explanation on this blog: https://timeseriesreasoning.com/contents/holt-winters-exponential-smoothing/

Then try to run the code, either on your own computer, or… Goto https://colab.research.google.com and get acquainted with Google Colab, a great place to try some serious analytics even using GPU or TPU acceleration. The code from the blogpost can be downloaded here: https://pdahlin.com/d?anpo_holtwinters (save this file and open in colab).

Can you reproduce the predictions? Do you think they are good?

Next, find similar timeseries data and try using that for prediction, following the same procedure as for the car sales!

Report in a way you find suitable.

This exercise is a good example of how we use analytic techniques and tools to get results without having to re-invent the wheel, so to say.