Forecasting is based on historical data. It means that if you would like to forecast potential future scenarios, you need to have historical data.

In this chapter, we wanted to cover how to prepare your historical data to create a Bitskout forecasting model.

There are three key components that historical data must have:

  1. Timestamp - it can be either an hour of the day, a day of the month or the month.

  2. Item identifier - if you predict sales, the item can be an SKU or serial number or a product name. For inventory forecasting, it can be a particular car class or an individual category of tools, or a certain toy.

  3. Value - this is a numerical value that we want to predict in the future. It can be total sales numbers, revenue, number of items in stock, or demand index. This value is always numerical.

Here are some examples of prepared data:

And this is an incorrect dataset - item identifier is missing.

Bitskout supports importing source data from your tool directly. Here is how the well-prepared source information looks like for one product on monday.com:

If there are two products that you want to predict the data for, the source data should have the following structure:

In order to test have a reference, here are some examples of datasets:

Did this answer your question?