Edit General Settings for a Model

Choose a validation type for your model. Select one of the following options:

• Training/Validation Ratio to randomly split the observations in the CRM Analytics dataset into a training set and a validation set. The default is 80% training and 20% validation, but you can adjust the slider to change the ratio as needed.
• Validation Dataset to specify a separate CRM Analytics dataset that contains the validation data. You see only the datasets that match the schema of your model’s dataset. Ideally, select a dataset that contains a diverse set of observations that reflect real-world conditions. If you want to compare models for accuracy, selecting the same validation dataset gives you a common, unbiased baseline for evaluation.
• None (the default) uses only k-fold validation.

You can choose among different algorithms for Einstein Discovery to use to create the model:

• GLM: Default. Generalized Linear Model is a regression-based algorithm.
• GBM: Gradient Boost Machine is a decision tree-based ensemble machine learning algorithm.
• XGBoost: A decision tree-based, ensemble machine learning algorithm.
• Random Forest: A supervised learning algorithm that uses multiple decision trees, randomization, and other optimization techniques.

Alternatively, select Model Tournament to have Einstein Discovery run all four algorithms and then show the results of the algorithm that performed best.