Using the Predictive Metric in Reports and Documents

With the predictive model implemented as a metric, it can be used in reports or documents to determine possible trends and outcomes. Creating a report or document for data mining is similar to creating a regular report or document. There are no special requirements, other than including the predictive metric. For steps to create a report, see the Basic Reporting Help. For steps to create a document, see the Document Creation Help

Predictive models are generated based on a certain level of input data. Therefore, in order for a model to produce valid results when scored, the model must be provided with the same level of input data. If the user of the model does not have access to this level of data, then the model cannot be expected to return valid scores. For example, a security filter on the user does not allow access to the appropriate level of data.

Since Exponential Smoothing gives greater weight to more recent data, time series models (see Creating a Predictive Model Using MicroStrategy) should be updated whenever new data is available. While this can be said for other types of model as well, this is particularly true for time series models since recent results can significantly influence the model. Therefore, for time series models, it is common to deploy the training metrics (see Creating a Predictive Model Using MicroStrategy) instead of the predictive metrics (see Creating a Predictive Model Using MicroStrategy). This guarantees the time series forecast results reflect the most recent data.