Common tasks in the Time Series Visualizer

These tasks are described in the sequence of the tabs, and on the tabs in alphabetical order.

Changing tabs

At the top of the page, the Time Series Visualizer provides the following tabs. Each of these tabs provides options for different sets of actions.

• Time Series
• Settings
• Publish
• Help

To access the options of a tab, click the appropriate tab.

Changing the time series to be displayed

If the model includes more than one time series, you can change the time series that are displayed in the Time Series Visualizer by clicking Select Time Series and selecting a different time series from the drop-down lists. The time series that is selected from the left drop-down list uses the left y-axis. The time series that is selected from the right drop-down list uses the right y-axis. If you change a time series in one of the views of the Time Series Visualizer, it is changed accordingly in the other views.

If you want to display only one time series curve in the visualizer, select None from one of the drop-down lists.

Changing views

The Time Series Visualizer provides the following views:

• Chart View
• Details View
• Chart and Details View

To display a view, click the appropriate view tab on the left side of the page.

Closing menus

You can close a menu by clicking the drop-down arrow or by clicking in the area to the right or to the left of the menu.

Closing the visualizer

You can close the visualizer by clicking the Close icon.

Displaying product information

You can display the product name and version information by clicking Help > Product Information.

Opening menus

You can open a menu by clicking its drop-down arrow.

Printing charts, tables, or summaries

By default, you can print the time series as a chart and a table including summary information about the time series by selecting Publish > Print. If you want to print only one of these items, clear the appropriate check boxes.

To print a chart, you can select whether to print the entire time series or only the part of the time series that is currently selected. You can also include the legend to be printed.

To print a table, you can select to print only selected rows or the entire table.

Retrieving help information

If you are connected to the Internet, you can retrieve help information for the Time Series Visualizer by selecting Help > Help topics.

Selecting forecast types and their confidence values

By default, for each time series, the curve of the best forecast and its confidence values are shown. You can add the curve of other forecast methods and the related confidence values by selecting the appropriate check boxes in the drop-down menu of the time series on the Time Series tab.

If you add the curve of a forecast method and its confidence value to the chart in the Chart View, the related columns are added to the table in the Details View.

To remove one of the confidence values or forecast methods, clear the appropriate check box.

Specifying the transparency of the original time series curve

By default, the value for the transparency of the original time series curve is set to 0. This means that the original time series curve is displayed with full saturation.

Larger values of transparency reduce the level of saturation. Valid values for the transparency are between 0 and 100. The curve disappears completely if the transparency is set to 100.

Tip: If you want to examine the moving average without being distracted by the original time series curve, type 100 in the appropriate entry field.

Specifying moving average

By default, moving averages are not displayed.

You might want to smooth the original time series curve to understand longer-term behaviour (trend) of your data without looking at short-term fluctuations. Short-term fluctuations are often meaningless for the business problem at hand or even random in nature. There might also be a shorter-term periodic behavior in your data that makes it more difficult to determine a trend. For example, consider unemployment data that typically vary with the seasons. In the northern hemisphere, there is often an increase in January even when the general trend is towards lower unemployment.

You can smooth a given curve by calculating for each time point in the curve the average value of all data points within a specified time window in the neighborhood of the time point. In the chart, the colors of the moving average curves and the original time series curves are the same.

To display the moving average of the original time series curve, follow these steps:

1. On the drop-down menu of the Time Series tab, select the Moving average check box for a particular time series.
2. From the list of methods, select one of the following methods to be used by clicking the drop-down arrow:

   Window centered

   In a centered window, the same period of time to the left and to the right of a given time point is used to calculate the average value. However, in a centered window, there are not enough time points available at the beginning and at the end of the original time series curve to accommodate for the selected window size. Therefore, average values cannot be calculated for the time points at the beginning and at the end of the time range.

   Window trailing

   In a trailing window, the time points to the left of a given time point (including the given data point) are used to calculate the average value. In a trailing window, there are not enough time points available at the beginning of the original time series curve to accommodate for the selected window size. Therefore an average value cannot be calculated for the data points at the beginning.

3. Modify the default value for the window size by typing values of your choice in the entry fields. You can also use the slider or click the spin buttons to specify the window size. By default, the window size is a quarter of the distance between the most left x-axis value and the most right x-axis value of the original time series data.

   Depending on the data type of the time series, the appropriate units are displayed in the drop-down menu. For example, if the time series data includes data types such as time, date, or timestamp, you can specify the window size by selecting time units such as milliseconds (ms), seconds (s), minutes (m), hours (h), days (D), months (M), or years (Y). If the time series includes data types such as integers or real data, you can specify the window size by specifying numbers.

   Note: If your original time series has periodic behavior, you must be careful how to set the window size.

   • If the goal is to remove periodic effects, it is recommended to use a multiple of the period length.
   • If the goal is to remove short-term fluctuations while preserving periodic behavior, you should use a value much smaller than the period length.

   In either case, you should avoid using values that are between multiples of the period length. Misleadingly, this might indicate periodic behavior that does not exist in the original data. This effect is particularly large near 1.5 period lengths, when the behavior seems to be inverse to the true behavior.

If the moving average curve is shown in the Chart View, the related column is added to the table in the Details View.