Test set beating training set


Test set beating training set



I have begun using RapidMiner recently and am having a strange problem with one of my workflows. I have split a dataset using the Split Data function, I have then built a random forest on the 90% Training set and applied that model on the 10% test set. However when I asses the performances, the test set consistently does better even as I vary the seeds. This result seems counter intuitive and I'm wondering if I have interpreted one of the arguments wrongly or am missing a detail? 


By the way I am aware that there are more efficient ways to set up this flow, I am trying alternative ways as a bit of practice 




Community Manager

Re: Test set beating training set

Ok, you have to be careful here with your setup because the results are misleading based on your choice of partition size for the Split operator. Why 90% and 10%? Why not 85% and 15%?  You will get varying results based on the size of your split, seed, and how you split the data. I noticed you used stratified sampling, which samples your data according to the class distribution of survivorship (Yes/No), so  you can get strange results there. 

What I suggest is to use a Cross Validation operator as your setup appears to try to mimic that thought process. I ran the process below by changing the seed and got that the Training Perf is slightly better than the Test Perf. Then I added a Cross Validation and measured the results there.


Also, you can use the Select Attributes operator to select the attributes you want inlieu of the R script if you want. 

<?xml version="1.0" encoding="UTF-8"?><process version="7.5.003">
  <operator activated="true" class="process" compatibility="7.5.003" expanded="true" name="Process">
    <parameter key="encoding" value="SYSTEM"/>
    <process expanded="true">
      <operator activated="true" class="retrieve" compatibility="7.5.003" expanded="true" height="68" name="Retrieve Titanic" width="90" x="45" y="34">
        <parameter key="repository_entry" value="//Samples/data/Titanic"/>
      <operator activated="false" class="r_scripting:execute_r" compatibility="7.2.000" expanded="true" height="68" name="Execute R" width="90" x="45" y="340">
        <parameter key="script" value="&#10;rm_main = function(data)&#10;{&#10;  return(data[,c(&quot;Sex&quot;, &quot;Passenger.Class&quot;, &quot;Age&quot;, &quot;Survived&quot;)])&#10;}&#10;"/>
      <operator activated="true" class="select_attributes" compatibility="7.5.003" expanded="true" height="82" name="Select Attributes" width="90" x="179" y="34">
        <parameter key="attribute_filter_type" value="subset"/>
        <parameter key="attributes" value="Passenger Class|Sex|Survived|Age"/>
      <operator activated="true" class="set_role" compatibility="7.5.003" expanded="true" height="82" name="Set Role" width="90" x="313" y="34">
        <parameter key="attribute_name" value="Survived"/>
        <parameter key="target_role" value="label"/>
        <list key="set_additional_roles"/>
      <operator activated="true" class="multiply" compatibility="7.5.003" expanded="true" height="103" name="Multiply" width="90" x="179" y="289"/>
      <operator activated="true" class="concurrency:cross_validation" compatibility="7.5.003" expanded="true" height="145" name="Validation" width="90" x="380" y="340">
        <parameter key="sampling_type" value="shuffled sampling"/>
        <process expanded="true">
          <operator activated="true" class="concurrency:parallel_random_forest" compatibility="7.5.003" expanded="true" height="82" name="RF for Xval" width="90" x="179" y="34">
            <parameter key="use_local_random_seed" value="true"/>
            <parameter key="local_random_seed" value="2110"/>
          <connect from_port="training set" to_op="RF for Xval" to_port="training set"/>
          <connect from_op="RF for Xval" from_port="model" to_port="model"/>
          <portSpacing port="source_training set" spacing="0"/>
          <portSpacing port="sink_model" spacing="0"/>
          <portSpacing port="sink_through 1" spacing="0"/>
          <description align="left" color="green" colored="true" height="113" resized="true" width="284" x="85" y="148">Builds a model on the current training data set (90 % of the data by default, 10 times).&lt;br&gt;&lt;br&gt;Make sure that you only put numerical attributes into a linear regression!</description>
        <process expanded="true">
          <operator activated="true" class="apply_model" compatibility="7.5.003" expanded="true" height="82" name="Apply Model (3)" width="90" x="45" y="34">
            <list key="application_parameters"/>
          <operator activated="true" class="performance" compatibility="7.5.003" expanded="true" height="82" name="Performance" width="90" x="179" y="34"/>
          <connect from_port="model" to_op="Apply Model (3)" to_port="model"/>
          <connect from_port="test set" to_op="Apply Model (3)" to_port="unlabelled data"/>
          <connect from_op="Apply Model (3)" from_port="labelled data" to_op="Performance" to_port="labelled data"/>
          <connect from_op="Performance" from_port="performance" to_port="performance 1"/>
          <connect from_op="Performance" from_port="example set" to_port="test set results"/>
          <portSpacing port="source_model" spacing="0"/>
          <portSpacing port="source_test set" spacing="0"/>
          <portSpacing port="source_through 1" spacing="0"/>
          <portSpacing port="sink_test set results" spacing="0"/>
          <portSpacing port="sink_performance 1" spacing="0"/>
          <portSpacing port="sink_performance 2" spacing="0"/>
          <description align="left" color="blue" colored="true" height="107" resized="true" width="333" x="28" y="139">Applies the model built from the training data set on the current test set (10 % by default).&lt;br/&gt;The Performance operator calculates performance indicators and sends them to the operator result.</description>
        <description align="center" color="transparent" colored="false" width="126">A cross validation including a linear regression.</description>
      <operator activated="true" class="split_data" compatibility="7.5.003" expanded="true" height="103" name="Split Data" width="90" x="380" y="187">
        <enumeration key="partitions">
          <parameter key="ratio" value="0.9"/>
          <parameter key="ratio" value="0.1"/>
        <parameter key="sampling_type" value="stratified sampling"/>
        <parameter key="use_local_random_seed" value="true"/>
        <parameter key="local_random_seed" value="2000"/>
      <operator activated="true" class="multiply" compatibility="7.5.003" expanded="true" height="103" name="Multiply (2)" width="90" x="514" y="34"/>
      <operator activated="true" class="concurrency:parallel_random_forest" compatibility="7.5.003" expanded="true" height="82" name="Random Forest" width="90" x="514" y="187">
        <parameter key="use_local_random_seed" value="true"/>
        <parameter key="local_random_seed" value="2110"/>
      <operator activated="true" class="multiply" compatibility="7.5.003" expanded="true" height="103" name="Multiply (3)" width="90" x="648" y="136"/>
      <operator activated="true" class="apply_model" compatibility="7.5.003" expanded="true" height="82" name="Apply Model (2)" width="90" x="715" y="34">
        <list key="application_parameters"/>
      <operator activated="true" class="performance_classification" compatibility="7.5.003" expanded="true" height="82" name="Trainig Perf" width="90" x="849" y="34">
        <list key="class_weights"/>
      <operator activated="true" class="apply_model" compatibility="7.5.003" expanded="true" height="82" name="Apply Model" width="90" x="782" y="289">
        <list key="application_parameters"/>
      <operator activated="true" class="performance_classification" compatibility="7.5.003" expanded="true" height="82" name="Test Perf" width="90" x="916" y="187">
        <list key="class_weights"/>
      <connect from_op="Retrieve Titanic" from_port="output" to_op="Select Attributes" to_port="example set input"/>
      <connect from_op="Select Attributes" from_port="example set output" to_op="Set Role" to_port="example set input"/>
      <connect from_op="Set Role" from_port="example set output" to_op="Multiply" to_port="input"/>
      <connect from_op="Multiply" from_port="output 1" to_op="Split Data" to_port="example set"/>
      <connect from_op="Multiply" from_port="output 2" to_op="Validation" to_port="example set"/>
      <connect from_op="Validation" from_port="performance 1" to_port="result 3"/>
      <connect from_op="Split Data" from_port="partition 1" to_op="Multiply (2)" to_port="input"/>
      <connect from_op="Split Data" from_port="partition 2" to_op="Apply Model" to_port="unlabelled data"/>
      <connect from_op="Multiply (2)" from_port="output 1" to_op="Random Forest" to_port="training set"/>
      <connect from_op="Multiply (2)" from_port="output 2" to_op="Apply Model (2)" to_port="unlabelled data"/>
      <connect from_op="Random Forest" from_port="model" to_op="Multiply (3)" to_port="input"/>
      <connect from_op="Multiply (3)" from_port="output 1" to_op="Apply Model (2)" to_port="model"/>
      <connect from_op="Multiply (3)" from_port="output 2" to_op="Apply Model" to_port="model"/>
      <connect from_op="Apply Model (2)" from_port="labelled data" to_op="Trainig Perf" to_port="labelled data"/>
      <connect from_op="Trainig Perf" from_port="performance" to_port="result 1"/>
      <connect from_op="Apply Model" from_port="labelled data" to_op="Test Perf" to_port="labelled data"/>
      <connect from_op="Test Perf" from_port="performance" to_port="result 2"/>
      <portSpacing port="source_input 1" spacing="0"/>
      <portSpacing port="sink_result 1" spacing="0"/>
      <portSpacing port="sink_result 2" spacing="273"/>
      <portSpacing port="sink_result 3" spacing="0"/>
      <portSpacing port="sink_result 4" spacing="0"/>



Thomas - Community Manager
LinkedIn: Thomas Ott

Re: Test set beating training set

Thanks for the response


Yes I am aware that some of my parameters were a bit weird but I was varying them all and getting similar results. Turns out I was changing the seed in the wrong operator, silly mistake. 


Regarding the choice of operators, I was just using things such as the execute R just to try out different operators and get a feel for how everything works. Thanks for the help Smiley Happy