Predictive modeling

Spotfire provides you with the tools to incorporate predictive models into your analysis using either regression modeling or classification modeling.

Note: The predictive modeling tools must be applied using the installed client.

Note: The predictive modeling tools uses data functions based on the Spotfire Enterprise Runtime for R (a/k/a TERR) to perform the calculations. Your Spotfire environment must support TERR data functions to be able to use the tools. See Usage of scripts and data functions for more information.

Regression modeling is useful for making numeric predictions, such as profit and expenses.
Classification modeling is useful for making predictions for typically two nodes or classes, such as whether a business transaction is fraudulent or legitimate.

The three tasks of predictive modeling include:

Fitting the model.
Evaluating the model.
Predicting from the model.

To fit the model, in the Regression modeling or Classification modeling dialog, select the model options and click OK. Spotfire Enterprise Runtime for R creates the model and returns it to the analysis. Each model creates a number of new data tables that you can use for further analysis.

A model page is created (see The model page) and the model is added to The Analytic models panel. You can include more than one model in your analysis, and then you can iterate and evaluate all models with new data.

After you have completed your evaluation, optionally you can predict from the model. When you include a model in the analysis, you can use the model to insert predicted columns into your data table and share the result with others.

Note: Columns of the data type Currency are converted to Real, which can result in a lack of precision in the modeling.