You might see an exception in the log files that Predictive Analytics Director is running out of the allocated memory. What is more, when using the Predictive Analytics Director portal, development of a model takes longer and the portal becomes unresponsive.

Cause

The data set that you used to build the model might be too big, or the amount of RAM that was allocated for PAD is not enough.

Solution

After you perform each troubleshooting step, check whether it solves your problem; if it does not, continue with the next step.

1. Set a smaller sample size and limit the number of predictors. You do not need a large data set to build a good predictive model.

   1. In the Data preparation step, set a smaller sample size. For more information, see Constructing a sample.

   2. In the Data analysis step, select a smaller number of predictors. For more information, see Defining the predictor role.

2. In the Data preparation step, reduce the number of fields by selecting the NOT USED type before they become predictors. For more information, see Constructing a sample.

3. If reducing the sample and the number of fields does not help, ask the server administrator to allocate more RAM for Predictive Analytics Director.

   You can use the Performance calculations spreadsheet to estimate RAM usage. The number might vary depending on the structure of the input data.