Actionable data to support or automate decision making is used by predictive and prescriptive analytics. Prediction of a quantity i.e. regression is required by many use cases. The actions based on the predictions is more important than the actual quantity predicted. Predictions of future events in the context of predictive and prescriptive analytics needs a full probability density distribution. The predictions might give some estimated value of expected fluctuations, but details will still be hidden in the full probability density distribution. In this case, advanced predictive models can be used to derive best point estimator for a given scenario.
Read more at: http://www.blue-yonder.com/blog-e/2015/07/20/predictions-as-a-probability-density-distribution/