Analytics Builder > Working with Predictive Models > Creating New Predictive Models > Add a Learner to a New Model
Add a Learner to a New Model
In Analytics Builder, a new learning technique can only be added to a model during advanced model configuration.
To add a learner technique to a new model:
1. Navigate to the Add Learner dialog box as follows:
◦ From the top of the Models list page, click New.
The New Predictive Model dialog box opens.
◦ Enter the necessary information on the Data Selection tab. (For more information, see Creating New Predictive Models.)
◦ Select the Advanced Model Configuration tab and click Add Learner.
The Add Learner dialog box opens.
2. Select one of the Learning Techniques listed on the left.
3. Provide values for the parameters displayed for the selected technique. The available parameters do not all apply to every learning technique. The following table describes each parameter and indicates which learners it applies to.
Parameter
Description
Learners it applies to
How many Learners do you want to include?
The number of learners that will be added to the model. If more than one, the system will clone the technique you are defining.
Linear Regression
Logistic Regression
Decision Tree
Support Vector Machines (SVM)
Neural Network
Random Forest
Gradient Boost
Maximum Depth
The maximum tree depth for each tree. Must be greater than 1.
Decision Tree
Gradient Boost
Random Forest
Number of Iterations
The number of iterations to use. Must be greater than 0.
Gradient Boost
Number of Trees
The number of trees to use. Must be greater than 1.
Random Forest
Hidden Unit %
The percentage of the number of input nodes to use in each hidden layer. Must be an integer greater than 0. Default value is 20.
Example: If the default value is used and the input layer has 10 nodes, each hidden layer will contain 10 * 0.2 = 2 hidden nodes.
* 
A minimum of 2 hidden nodes per hidden layer is required. If the input layer has 5 nodes and the default hidden unit percentage is used, 5 * 0.2 = 1. In this case, the number of hidden nodes on each hidden layer will be 2 instead of 1.
Neural Network
Layer Count
The number of layers to use in the neural net. Must be an integer greater than or equal to 2. Default value is 3. A value of 2 means 1 input layer and 1 output. A value of 3 adds a hidden layer and a value of 4 adds a second hidden layer.
Neural Network
Hidden Layer Activation Function
The type of output transformation to apply to nodes in the hidden layers so that values are maintained within a manageable range. Valid values include Sigmoid, TanH (hyperbolic tangent), and ReLU (rectified linear units). (Default is Sigmoid)
Neural Network
Output Layer Activation Function
The type of output transformation to apply to nodes in the output layer so that values are maintained within a manageable range. Valid values include Linear and Sigmoid. If no value is specified, an activation function is automatically selected based on the OpType of the goal variable.
Neural Network
For more information about each learner, see Learners and Ensemble Techniques.
4. Click Add. The new learner is added to the model configuration.
Was this helpful?