Assign Data to Clusters

Your workspace should contain a trained clustering model. You can create and train a clustering model by using either of these methods:

• Configure the K-means algorithm using the K-Means Clustering module, and then train the model using a dataset and the Train Clustering Model module.

• Configure a range of options for the K-means algorithm using K-Means Clustering and then train the model using the Sweep Clustering module.

• Add an existing trained clustering model from the Saved Models group in your workspace.

Attach the trained model to the left input port of Assign to Clusters (deprecated).

Attach a new dataset as input.

Labels are optional in the dataset. Generally, clustering is an unsupervised learning method so it is not expected that you will know categories in advance.

By default, all columns in the input dataset are returned in the results. If you want to use fewer columns when creating cluster predictions, use Select Columns in Dataset to select a subset of the columns. However, the input columns must be the same as the columns that were used in training the clustering model, or an error will occur.

Leave the option Check for Append or Uncheck for Result Only selected if you want the results to contain the full input dataset, together with a column indicating the results (cluster assignments).

If you deselect this option, you get back just the results. This might be useful when creating predictions as part of a web service.

Run the experiment.

The Assign to Clusters (deprecated) module returns two types of results on the Results dataset output:

• Click the output of the module and select Visualize to view a PCA graph or Principal Components Graph.

  In a PCA graph, the collection of values in each cluster is mapped to two component axes: the first component axis is the combined set of features that captures the most variance in the model, and is plotted on the x-axis (Principal Component 1). The next component axis represents some combined set of features that is orthogonal to the first component and that adds the next most information to the chart, plotted on the y-axis (Principal Component 2). From this graph, you can see the maximum separation that can be attained between the clusters, and how all clusters are distributed along the axes that represent the principal distinguishing components.

• To view the table of results for each case in the input data, you can attach the Convert to Dataset module to view the results in Machine Learning Studio.

  The results dataset contains the cluster assignments for each case, and a distance metric that gives you some indication of how close this particular case is to the center of the cluster.

  Output column name	Description
  Assignments	A 0-based index that indicates which cluster the data point was assigned to.
  DistancesToClusterCenter no. n	For each data point, this value indicates the distance from the data point to the center of the assigned cluster, and the distance to other clusters. The metric used to calculate distance is determined when you configure the K-means clustering model.