Model requires a pair of training and test tables, with input features and a response or target field, one record for each entity being modeled. Typically, the training table contains 80% of the records, with the remaining 20% applied to the test table to validate the accuracy of the model.
Model automatically filters some tables, based on their name. Tables that contain the following case-sensitive strings in their name cannot be selected as a training or test tables:
- _MODEL
- _MONITOR
- _MODEL_PREDICTIONS
- _COUNT_OUTPUT
- _STAT_OUTPUT
- _ACC_OUTPUT
The structure for the Model training and test tables includes two required fields and any attributes applicable the model.
The two required fields are:
- entity reference
- response
Sample Model Profile Table
This sample table contains model input data that could be used to create an upgrade offer or churn prediction model for telecommunications customers.