Configuration

This section explains each of the properties in the configuration blade.

General

PropertyDescription

Collection

This associates the Agent to a specific Collection (default to that of the current Data Stream).

Refer to Collections and Stream Hosts to understand more about Collections.

Training

PropertyDescription

Learning Algorithm

The algorithm used to train the model.

Dataset

The training file used to train the algorithm.

Has Header

Tick to auto-populate the feature names from the first row of the training file, or the features are named column0, column1, etc.

Separator Character

The character used to separate data values in the training file.

Features (Name, Type, and Variable Type)

The grid of features is categorized by name, data type, and variable type.

  • Name: Feature name

  • Type: Data type of the feature

  • Variable Type: options are feature (default), exclude (ignore certain features from training), or class variable (the output).

Algorithm Parameters

The following list contains definitions of all parameters, but the combination of parameters used by each algorithm differs.

PropertyDescription

L1 Regularization

L1 penalty equal to the absolute value of the magnitude of coefficients.

L2 Regularization

L2 penalty which is equal to the square of the magnitude of coefficients.

Learning Rate

The amount of change in model in each iteration for better prediction.

Number of Interations

The number of times algorithms parameters are updated in a single run.

Number of Trees

The total number of trees generated on training data.

Number of Leaves

The number of leaves per tree in the model.

Minimum Example Count Per Leaf

The minimum number of samples required to be at a leaf node.

Model Options

PropertyDescription

Cross-Validation Folds

Indicates the number of times the fitting procedure is executed on the training data.

Deterministic Seed

Set this to a whole number for repeatable/deterministic results across multiple trainings.

Before configuring the Agent, please ensure that its input endpoint is connected to a parent Agent which will be sending data to it.

Input Mapping

Each parent Agent attribute is mapped against a feature in the training file so that test data matches the training data.

Refer to Setup Input Mappings for step-by-step instructions.

Endpoints

NameDescription

Input

This endpoint is used to receive data from the parent Agent.

Output

Events received from the parent Agent are assigned a class value and made available to this endpoint.

Last updated