Viewing a Mining Model with the Microsoft Time Series Viewer
The Microsoft Time Series Viewer in Microsoft SQL Server 2005 Analysis Services (SSAS) displays mining models that are built with the Microsoft Time Series algorithm. The Microsoft Time Series algorithm is a regression algorithm for use in creating data mining models to predict continuous columns, such as product sales, in a forecasting scenario. For more information about this algorithm, see Microsoft Time Series Algorithm.
When you browse a mining model in Analysis Services, the model is displayed on the Mining Model Viewer tab of Data Mining Designer, using the appropriate viewer for the model.
The Microsoft Time Series Viewer provides the following tabs:
Click the Decision Tree tab to display the tree view of the time series model. Click the Chart tab to display plots of the time series in a chart that includes historical data and predicted data.
When you build a decision tree model, Analysis Services builds a separate tree for each continuous predictable attribute. You can view an individual tree by selecting it from the Tree list on the Decision Tree tab of the viewer.
A decision tree is composed of a series of splits, with the most important split, as determined by the algorithm, at the left of the viewer in the All node. Additional splits occur to the right. The split in the All node is most important because it contains the strongest split-causing conditional in the dataset, and therefore it caused the first split.
You can expand or collapse individual nodes in the tree to show or hide the splits that occur after each node. You can also use the options on the Decision Tree tab to affect how the tree is displayed. Use the Show Level slider to adjust the number of levels that are shown in the tree. Use Default Expansion to set the default number of levels that are displayed for all trees in the model.
The viewer displays a regression formula at each leaf in the tree. The regression formula is used to predict values in the time series. Intermediate nodes contain the condition that caused a split and support for the split. The viewer also displays a diamond chart for each node in the tree. The diamond chart has a line that represents the range of the attribute. The diamond is located at the mean for the node, and the width of the diamond represents the variance of the attribute at that node. A thinner diamond indicates that the node can create a more accurate prediction.
The shading of the background color for each node signifies the number of cases that exist in the node. To find the exact number of cases in a node, pause the pointer over the node to view an InfoTip for the node.
The Charts tab displays the behavior of a time series, together with the predicted values over time. The vertical axis of the chart represents the value of the series, and the horizontal axis represents time.
Use the Abs button to toggle between absolute and relative curves. If your chart contains multiple models, the scale of the data for each model may be greatly different. If you leave the toggle on absolute, one model will show a flat line, while other model show dramatic changes. This occurs because the scale of one model is so much larger than the scale of other models that its changes are minimized. Setting the toggle to relative changes the scale to a percent of change, and therefore shows the changes in each model on a relative scale.
If the mining model contains multiple time series, choose the series to display in the chart by selecting the corresponding sets in the list to the right of the viewer. You can then select the individual series to display by selecting the appropriate check boxes in the legend.
The chart displays both historical and future data. Future data is shaded, to differentiate it from historical data.
You can adjust the range of time that is displayed by using the zoom options. You can also view a specific time range by clicking the chart, dragging a time selection across the chart, and then clicking again to zoom in on the selected range.
You can select how many future time steps you want to see in the model by using Prediction Steps. The Show Deviations check box adds error bars so that you can see how accurate the predicted value is.