Time for action – predicting future sales with Microsoft Time Series

In this example, we use a dataset to train a Time Series algorithm with sales information of previous years and then we will see how a Time Series algorithm will help to predict future sales based on the trained pattern.

1. Open Adventure Works DW2012.dsv and add a new table to it. Choose vTimeSeries from the list of tables.

2. Right-click on the vTimeSeries table in DSV and explore the data. You will see that this table contains the sales amount and quantity based on the product model and region and month.

3. In the DSV designer, select both the TimeIndex and ModelRegion columns from vTimeSeries and right-click and choose Set Logical Primary Key.

4. Create a new mining structure from the existing relational database and choose Microsoft Time Series as the mining model. Set the vTimeSeries table as Case table.

5. In the Specify the Training Data window, set TimeIndex and ModelRegion as the Key columns and set Amount and Quantity as the Input...