"Using a Kmeans model (clm file) created w/ a sample to cluster my population"

KeithrKeithr Member Posts: 10 Contributor II
edited May 23 in Help
I created a kmeans model with 7 clusters using a sample of 10K records.  I now want to cluster my whole population (~ 1 mill records) and am using the operators ClusterModelReader to read the cluster model I created w. the sample (clm file) and the operator ClusterModel2ExampleSet to cluster the entire population. 

Your description for this operator states "This Operator clusters an exampleset given a cluster model. If an exampleSet does not contain id attributes it is probably not the same as the cluster model has been created on. Since cluster models depend on a static nature of the id attributes, the outcome on another exampleset with different values but same ids will be unpredictable.".  Does this mean that it will only cluster the records that I used to create the model, and will not do any new records? 

The process below finishes correctly but only clustered the records that had been clustered in the sample file.  All other records had a blank cluster # in the output file.

Is there a way to use the model I created to cluster new records or do I have to run the kmeans algorithm on the 1 mill record file and not use the clm file created from the sample data?

Thanks in advance.

Keith


      <operator name="ClusterModelReader" class="ClusterModelReader">
          <description text="The cluster model 8051_Lifestyle_Matches_Excel.clm is the exact model I used for the Excel study so use it to cluster the population"/>
          <parameter key="cluster_model_file" value="C:\Documents and Settings\krobinson\My Documents\rm_workspace\Clustering\8051_Lifestyle_Matches_Excel.clm"/>
      </operator>
      <operator name="ClusterModel2ExampleSet" class="ClusterModel2ExampleSet">
      </operator>
      <operator name="OperatorChain" class="OperatorChain" expanded="yes">
          <operator name="PSVExampleSetWriter" class="CSVExampleSetWriter">
              <parameter key="column_separator" value="|"/>
              <parameter key="csv_file" value="C:\Documents and Settings\krobinson\My Documents\rm_workspace\Clustering\8051_Population_Lifestyle.psv"/>
          </operator>
      </operator>

Answers

  • steffensteffen Member Posts: 347  Guru
    Hello all of you

    I wanted to try something similar this weekend ... and...

    I checked the code, but indeed yes, K-Means and Clustermodels in general are not easy applicable (as far as I see) to new data :(.

    There are clustering algorithms out there which have no predefined way of cluster a new item, DBScan for instance (and so are "static"). But a centroid-oriented algorithm like k-means, whose clustering strategy implies how to assign a cluster to a new item ...

    Well, lets wait for statement from the official side

  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,525   Unicorn
    Hi,
    here's the official side :) I was **** sure having implemented this features for centroid cluster models. Maybe it got lost during the restructuring of the new cluster plugin's model class structure.
    I will add it these days in the developer branch and get back to you, if I'm finished.

    Greetings,
      Sebastian
  • KeithrKeithr Member Posts: 10 Contributor II
    Hi,

    So just to make sure I understand, Kmeans should be able to cluster new items based on the model created with the sample data, and I used the correct operators (clusterModelReader/ClusterModel2ExampleSet), right?

    If this is the case I'll wait for your update.

    Thanks for the quick response.

    Keith
  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,525   Unicorn
    Hi Keith,
    I have commited the new feature into the developer branch. If it works stable we will merge it down to the stable version and include it into the next enterprise update.

    Greetings,
      Sebastian
  • MuehliManMuehliMan Member Posts: 85  Guru
    Hi,

    I have a problem with clustering too. With my preprocessed Source I do:
       
    ...
    <operator name="Datenset Clustern" class="OperatorChain" expanded="yes">
            <operator name="KMeans" class="KMeans">
            </operator>
            <operator name="ClusterModel2ExampleSet" class="ClusterModel2ExampleSet">
            </operator>
            <operator name="LinearRegression" class="LinearRegression">
                <parameter key="keep_example_set" value="true"/>
            </operator>
        </operator>
    I expected at least 2 models as output, but I get one with cluster column. At least I can't see the two Example Sets and I found no operator that splits  the ExampleSet according to value.
    What I want to do is basically apply a e.g. LinearRegression Operator to all of the Clusters individually.

    Thx in advance.
    Markus
  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,525   Unicorn
    Hi Markus,
    this is possible using a combination of iterators and filtering:
    <operator name="Root" class="Process" expanded="yes">
        <operator name="ExampleSetGenerator" class="ExampleSetGenerator">
            <parameter key="number_of_attributes" value="2"/>
            <parameter key="target_function" value="global and local models classification"/>
        </operator>
        <operator name="Datenset Clustern" class="OperatorChain" expanded="yes">
            <operator name="KMeans" class="KMeans">
            </operator>
            <operator name="ValueIterator" class="ValueIterator" expanded="yes">
                <parameter key="attribute" value="cluster"/>
                <operator name="ExampleFilter" class="ExampleFilter">
                    <parameter key="condition_class" value="attribute_value_filter"/>
                    <parameter key="parameter_string" value="cluster = %{loop_value}"/>
                </operator>
                <operator name="LinearRegression" class="LinearRegression" breakpoints="after">
                    <parameter key="keep_example_set" value="true"/>
                </operator>
                <operator name="ModelWriter" class="ModelWriter">
                    <parameter key="model_file" value="C:\Dokumente und Einstellungen\sland\Eigene Dateien\yale\workspace\ClusteredLR - %{loop_value}.mod"/>
                </operator>
            </operator>
        </operator>
    </operator>
    Greetings,
      Sebastian
  • MuehliManMuehliMan Member Posts: 85  Guru
    Hi Sebastian,

    that was exactly what I was searching for!

    Thank you very much!
    Markus
Sign In or Register to comment.