Chat now with support
Chat with Support

Spotlight on Oracle 11.0 - Getting Started Guide

Welcome to Spotlight Install Spotlight Start Spotlight Spotlight on Oracle Spotlight on Oracle Data Guard Spotlight on Oracle RAC Spotlight on Unix Spotlight on Windows Troubleshooting: Connection Problems

Prediction Interval

In Predictive Diagnostics charts, a prediction interval is used to indicate the range where the future values of a specified Cost Metric or Bottlenecks are expected to lie. The sample chart below shows such a prediction interval.

The prediction interval is generated from the (blue) raw data points, and indicates the range where future values are expected to lie.

  • The prediction line, the central line, represents the model that best fits the raw data points. Assuming that the system continues to change in a consistent manner, any future data points are expected to occur around this line.
  • The outer boundary lines border the prediction interval, and represent the confidence level for the chart. By default, they enclose 95% of the raw data points collected for the database; you can set the confidence level set in Predictive Diagnostics - View | Options | Predictions General Configuration.
  • The orange region is the prediction interval, the region where the future values of a specified Cost Metric or Bottlenecks are expected to lie. Since most of the future data points SHOULD be closer to the prediction line than the outer lines, we also need to indicate WHERE future points are likely to reside within the prediction interval. We do this via the color density of the orange band.

    The color density is NOT uniform — the band is most dense near the prediction line, where we expect most future values to occur, and less dense at the outer boundaries, where fewer data points will occur. If the boundary lines enclose 95% of the raw data points, we expect 68% of the future data points to reside in the central half of the prediction interval.

Flaring of the Prediction Interval

The flaring of the prediction interval with time indicates that the further into the future that predictions are made, the less accurate they will be. The amount of flare is determined by how closely the raw data points cluster around the prediction line.

 

Related Documents

The document was helpful.

Select Rating

I easily found the information I needed.

Select Rating