Forecasting time series

This tutorial introduces the Time Series mining function. The sample data contains records about the ticket sales of two airlines. The Time Series mining function is used to forecast values that are measured over time.

The database DWESAMP needs to be prepared on the database server. See Preparing the data for the tutorials.

This tutorial shows how to use the Time Series mining function to forecast time series based on sample data in the DWESAMP database. The sample data describes ticket sales of airlines. It is included in the table AIRLINE in the schema TIMESERIES.

Create a new data warehousing project:
1. From the menu of the Design Studio, click File > New > Data Warehousing Project.
2. On the Data Warehousing Project page of the New Project wizard, type Time Series tutorial in the Project name entry field and click Finish.
Create a new mining flow to forecast time series for airline ticket sales in the Time Series tutorial project:
1. In the Data Project Explorer, expand the project Time Series tutorial, right-click the folder Mining Flows, and select New > Mining Flow from the popup menu.
2. On the Data Mining Flow page of the New Data Flow wizard, type Forecasting Ticket Sales in the entry field, select to work against a database, and click Next.
3. On the Select Connection page of the New Data Flow wizard, select DWESAMP from the list of database connections and click Finish.
Add operators to the mining editor canvas and connect them:
1. In the mining flow editor, place a Table Source operator on the canvas.
2. On the Select Database Table dialog, expand the schema TIMESERIES, select the table AIRLINE, and click Finish.
3. In the mining flow editor, place the Time Series operator on the canvas to the right of the Table Source operator.
4. In the Table Source operator and the Time Series operator, click the Plus sign to display the columns of the AIRLINE table.
  The AIRLINE table includes data about ticket sales of two different airlines. The ticket sales are counted per month.
5. In the mining editor, place the Visualizer operator on the canvas to the right of the Time Series operator.
6. Connect the Table Source operator with the Time Series operator by sequentially clicking the output port of the Table Source operator and the Input port of the Time Series operator.
7. Connect the Time Series operator with the Visualizer operator by sequentially clicking the Model port of the Time Series operator and the Model port of the Visualizer operator.
Edit properties of the Time Series operator:
1. On the mining editor canvas, select the Time Series operator.
  In the Properties view below the mining editor canvas, the properties of the selected operator are displayed.
2. In the Properties view, click Model Name and replace the default name with Airline Tickets Sold.
3. In the Properties view, click Column Properties:
  1. In the list of available columns, select TIME and move it to the Time column by clicking the Right arrow.
  2. In the list of available columns, select the columns PASSENGERS_AB and PASSENGERS_BC and move them to the Value columns by clicking the Right arrow.
4. To save the updates, click Finish.
Start the mining flow:
1. On the tool bar of the Design Studio, click the icon to start the mining flow.
2. On the Flow Execution dialog, keep the default settings and click Execute.
  The Mining Flow Execution Status window is opened. When the mining run is completed, the Chart view of the Time Series Visualizer is opened.
Browse the result of the Time Series mining function in the Time Series visualizer. By default, the Chart view is displayed.
1. Open the Details view by clicking the appropriate tab to browse the result in table format.
2. Open the Chart and Details view for quick reference in the chart and in the table.
3. In the Details view of the Chart and Details view, click Summary to browse a summary of the result.
4. Zoom the graph by clicking the Zoom icon at the bottom of the visualizer.
5. Move the graph by activating the Hand tool at the bottom of the visualizer.

To summarize, this tutorial showed you how to create a new data warehousing project, create a mining flow to forecast time series, edit properties of a Time Series operator, and to browse the result in the Time Series Visualizer.

Feedback