Adaptive Model form
Completing the Memory estimation tab

Use this tab to estimate the memory (RAM) that the Adaptive Decision Manager (ADM) system needs to accommodate the models using the adaptive model configuration. The calculation applies linear growth for the number of models that you expect to generate in the ADM system, and quadratic growth for the number of predictors that you are using, or expecting to use, in the adaptive model configuration. To get an overall calculation, combine the estimated memory usage of the adaptive model configurations that is used to generate models in the same ADM server. You can use the memory estimation to tune server memory requirements and validate whether the adaptive model rule remains within the existing memory constraints.

Note: Any changes that you make to the values in this tab are reset to default ones when you refresh or reopen the rule.

1. In the NUMBER OF ADAPTIVE MODELS field, specify the number of models that you expect to be generated based on the adaptive model configuration.

2. In the NUMBER OF PREDICTORS column, specify the number of numeric and symbolic predictors.

   You can test different values to perform exploratory analysis or anticipate changes in learning data. If the number of predictors is different from the number of predictors added in the Predictors tab, a warning message appears.

3. In the AVERAGE NUMBER OF UNIQUE VALUES column, specify the average number of unique values the numeric and symbolic predictors can have.

   • The average number of unique numeric values is between 1 and 800; by default, it is 100.

   • The average number of unique symbolic values is between 1 and 200; by default, it is 15.

4. In the AVERAGE LENGTH column, specify the average length for symbolic values. By default, it is 25.

5. Click Estimate and check the Estimated memory usage.

Related information

Open topic with navigation