WebSphere Business Space powered by WebSphere, Version 6.2.0


Collecting history and prediction data

Use the following procedures to configure history and prediction capabilities for a KPI using the KPI Manager widget.

KPI History

You can track and analyze KPI changes over time. Snapshots of KPI values can be persisted so that you can view and analyze the historical values.
  1. Open KPI Manager and select a business process model. A list of KPIs is displayed.
  2. Select a KPI and select Actions > History and Prediction. The KPI History and Prediction window opens. From here, you can see how the KPI is configured for collecting history and prediction data.
  3. Under KPI History is the Track historical data for this KPI option. Select this option to enable historical data collection. Updated hourly means that the historical data collected is uploaded to the server on an hourly basis.
  4. Under Calculate KPI history data retroactively, you can instantaneously compute past data for specific periods of time. For example, you want to compute the existing data for this KPI for the past 3 months. Select Three months from the list and click Calculate.
  5. To replace any history data that was previously collected (if applicable) with the history data for the period that are about to calculate, select the Replace existing history data with this information option. When the Replace existing history data with this information option is not enabled, the information that you are about to calculate is added to the history collection, in addition to any existing data.
  6. When a confirmation message is displayed, click OK. History for the period that you selected is being generated in the background.

KPI Prediction

The KPI prediction capability is dependant upon the historical data that has been collected for the KPI. After history tracking is enabled, you can then configure the prediction options.

Under KPI Prediction, you can get started using the prediction feature by either selecting a basic option or by creating a prediction model.

To see prediction information about the KPI that is based only on the historical data that has been collected, select Enable KPI Prediction based on the historical data for this KPI and then click OK. To view the prediction chart, go to the KPI History and Prediction.

Creating a prediction model
  1. Click New Model to create a prediction model for this KPI. The Prediction Model Definition window opens.
  2. Enter information for the prediction model:
    1. Enter a name for the model in the Model name field.*
    2. Under Prediction data points, select the granularity of data points on the graph. For example, Once an hour means that the data points on the graph will be hourly.
    3. For Prediction period, enter the number of periods for the prediction, or select the To the end of the KPI time period option. For example, 11 and Once an hour means 11 hours. When the To the end of the KPI time period option is not available, it means that the KPI definition does not have a time period configured.
    4. Optional: Enable the Include a repeating business pattern option if you want to capture any repeating patterns (when applicable). This number is used in the formula in calculating prediction. For example, a salon owner notices that the salon draws more people during the weekend than during the week. There is a pattern in this business model's number of customers, where, during the first 5 days of the week, the customer number is low, and during the weekend, the number rises. In this example, the repeating business pattern is every week (or every 7 days).
    5. Click OK to save the prediction model.

* Limitations in this window could prevent backspacing and deleting in some fields.




Support | Terms of use | Contact IBM
Copyright IBM Corporation 2007, 2008. All Rights Reserved.
This information center is powered by Eclipse technology. (http://www.eclipse.org)