Merge Count Transform
Updated: April 12, 2016
Creates a set of features based on a counts table
You can use the Merge Count Transform module to combine two sets of count-based features. By merging two sets of related counts and features, you can potentially improve the coverage and distribution of the features.
In general, learning from counts is particularly useful in large data sets with high-cardinality features. Therefore, the ability to combine multiple datasets into count-based feature sets without having to reprocess the datasets makes it easier to gather statistics on very large datasets and apply them to new datasets. For example, count tables can be used to collect information over terabytes of data. You can re-use those statistics to improve the accuracy of predictive models on small data sets.
To merge two sets of count-based features, the features must have been created using tables that have the same schema: that is, both sets must use the same columns, with the same names and data types.
To use Merge Count Transform, you must have created at least one count-based transformation, and made it available in your workspace. For saved count-based transformations, look in the Transforms group. For transformations in th current experiment, see the outputs of the following modules:
Build Counting Transform. Creates a new count-based transformation from source data.
Modify Count Table Parameters. Takes an existing count transformation as an input and outputs an updated transformation.
Import Count Table. This module supports backward compatibility with older experiments that used count-based learning. If you used Import Count Table to analyze the distribution of values in a dataset, and then converted the values to features using the Count Featurizer (deprecated) module, use Import Count Table to convert the results to a transformation.
Add the Merge Count Transform module to the experiment, and connect a transformation to each input.
Tip The second transformation is an optional input – you can connect the same transformation twice, or connect nothing on the second input port.
If you do not want the second dataset to be weighted equally with the first, specify a value for Decay factor. The value that you type indicates how the set of features from the second transformation should be weighted.
For example, the default value of 1 weights both sets of features equally. A value of .5 means that the features in the second set would have half the weight of those in the first set.
Optionally, add an instance of the Apply Transformation module, and apply the transformation to a dataset.
You can see examples of how this module is used by exploring these sample experiments in the Model Gallery:
The Learning with Counts: Binary Classification sample demonstrates how to use the learning with counts modules to generate features from columns of categorical values for a binary classification model.
The Learning with Counts: Multiclass classification with NYC taxi data sample demonstrates how to use the learning with counts modules for performing multiclass classification on the publicly available NYC taxi dataset. The sample uses a multiclass logistic regression learner to model this problem.
The Learning with Counts: Binary classification with NYC taxi data sample demonstrates how to use the learning with counts modules for performing binary classification on the publicly available NYC taxi dataset. The sample uses a two-class logistic regression learner to model this problem.
Name | Type | Description |
|---|---|---|
Previous counting transform | The counting transform to edit. | |
New counting transform | The counting transform to add. |
Name | ToHide | Type | Range | Optional | Description | Default |
|---|---|---|---|---|---|---|
Decay factor | dampingFactor | Float | Required | 1.0f | The decay factor to be multiplied to the counting transform at the right input port. |
Name | Type | Description |
|---|---|---|
Merged counting transform | The merged transform. |
Exception | Description |
|---|---|
Exception occurs if one or more of inputs are null or empty. | |
Exception occurs when a counting transform is invalid. |