Time-series Forecast Model - Remove data points

Last but not least, besides adding country holidays and special events to your model, you can also delete some data points from your dataset.

As we saw in the previous post in the second example, we handled the begging of June 2020, the start of a promotion, as a special event. But in case that we don't know anything about that period, we come to a problem where the model carries the influence of that period to all the same dates in the past and in the future.

In Graphite, all you have to do is enter the start time and the end time of the period you want to delete. In case your time period lasts just one day, the start and the end date should be the same. Also, you can remove multiple data periods. It is important to give the model as much information as possible in order to get the most accurate prediction. In order to do that, you have to know your data, what type of data you are managing, on which date you need to pay attention to, which days are outliers, etc.