undersampling the large class
PP
New Altair Community Member
Answers
-
Split the original dataset ( Positive vs Negative class). Keep all cases of the rare class and sample from the frequent class.
Here's some code that undersamples the frequent class (class=yes) :<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<process version="6.5.002">
<context>
<input/>
<output/>
<macros/>
</context>
<operator activated="true" class="process" compatibility="6.5.002" expanded="true" name="Process">
<process expanded="true">
<operator activated="true" class="retrieve" compatibility="6.5.002" expanded="true" height="60" name="Retrieve Golf" width="90" x="45" y="75">
<parameter key="repository_entry" value="//Samples/data/Golf"/>
</operator>
<operator activated="true" class="multiply" compatibility="6.5.002" expanded="true" height="94" name="Multiply" width="90" x="179" y="75"/>
<operator activated="true" class="filter_examples" compatibility="6.5.002" expanded="true" height="94" name="Filter Examples (2)" width="90" x="313" y="210">
<list key="filters_list">
<parameter key="filters_entry_key" value="Play.equals.no"/>
</list>
</operator>
<operator activated="true" class="filter_examples" compatibility="6.5.002" expanded="true" height="94" name="Filter Examples" width="90" x="313" y="30">
<list key="filters_list">
<parameter key="filters_entry_key" value="Play.equals.yes"/>
</list>
</operator>
<operator activated="true" class="sample" compatibility="6.5.002" expanded="true" height="76" name="Sample" width="90" x="447" y="30">
<parameter key="sample_size" value="5"/>
<list key="sample_size_per_class"/>
<list key="sample_ratio_per_class"/>
<list key="sample_probability_per_class"/>
</operator>
<operator activated="true" class="append" compatibility="6.5.002" expanded="true" height="94" name="Append" width="90" x="514" y="120"/>
<connect from_op="Retrieve Golf" from_port="output" to_op="Multiply" to_port="input"/>
<connect from_op="Multiply" from_port="output 1" to_op="Filter Examples" to_port="example set input"/>
<connect from_op="Multiply" from_port="output 2" to_op="Filter Examples (2)" to_port="example set input"/>
<connect from_op="Filter Examples (2)" from_port="example set output" to_op="Append" to_port="example set 2"/>
<connect from_op="Filter Examples" from_port="example set output" to_op="Sample" to_port="example set input"/>
<connect from_op="Sample" from_port="example set output" to_op="Append" to_port="example set 1"/>
<connect from_op="Append" from_port="merged set" to_port="result 1"/>
<portSpacing port="source_input 1" spacing="0"/>
<portSpacing port="sink_result 1" spacing="0"/>
<portSpacing port="sink_result 2" spacing="0"/>
</process>
</operator>
</process>0 -
Many tanks for your help. I did this technique but my yes prediction in the confusion matrix is only 30%. I need better technique for over/under sampling.
Paulo Praca0 -
are you sure that this is because of class balance?0
-
No I'm not sure, I'm a civil engineer and i have a poor knowledge in data mining. In my municipality the water and sewer services produce in a daily basis information about the networks, and I tink data mining techniques could be a key to better understand why and where the failures occurred.
I could send you my example data if you are interested.
Thanks for your answer,
Paulo Praça0 -
<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<process version="7.0.001">
<context>
<input/>
<output/>
<macros/>
</context>
<operator activated="true" class="process" compatibility="7.0.001" expanded="true" name="Process">
<process expanded="true">
<operator activated="true" class="retrieve" compatibility="7.0.001" expanded="true" height="68" name="Retrieve Desentupimentos" width="90" x="45" y="187">
<parameter key="repository_entry" value="//Local Repository/data/Desentupimentos"/>
</operator>
<operator activated="true" class="set_role" compatibility="7.0.001" expanded="true" height="82" name="Set Role" width="90" x="179" y="187">
<parameter key="attribute_name" value="Obstrucoes"/>
<parameter key="target_role" value="label"/>
<list key="set_additional_roles"/>
</operator>
<operator activated="true" class="select_attributes" compatibility="7.0.001" expanded="true" height="82" name="Select Attributes" width="90" x="313" y="187">
<parameter key="attribute_filter_type" value="subset"/>
<parameter key="attributes" value="Obstrucoes|ANO_INSTALACAO|COD_MATERIAL|COMP|SECCAO|SISTEMA|TIPO_REDE"/>
</operator>
<operator activated="true" class="filter_examples" compatibility="7.0.001" expanded="true" height="103" name="Filter Examples" width="90" x="447" y="187">
<parameter key="parameter_expression" value="! ((missing([ANO_INSTALACAO])))"/>
<parameter key="condition_class" value="expression"/>
<list key="filters_list">
<parameter key="filters_entry_key" value="COD_MATERIAL.does_not_equal.NC"/>
</list>
</operator>
<operator activated="true" class="filter_examples" compatibility="7.0.001" expanded="true" height="103" name="Filter Examples (2)" width="90" x="581" y="187">
<list key="filters_list">
<parameter key="filters_entry_key" value="SECCAO.ne.0\.0"/>
</list>
</operator>
<operator activated="true" class="filter_examples" compatibility="7.0.001" expanded="true" height="103" name="Filter Examples (3)" width="90" x="45" y="289">
<list key="filters_list">
<parameter key="filters_entry_key" value="COD_MATERIAL.does_not_equal.NC"/>
</list>
</operator>
<operator activated="true" class="filter_examples" compatibility="7.0.001" expanded="true" height="103" name="Filter Examples (4)" width="90" x="179" y="289">
<parameter key="parameter_expression" value="COMP>=10"/>
<parameter key="condition_class" value="expression"/>
<list key="filters_list"/>
</operator>
<operator activated="true" class="split_data" compatibility="7.0.001" expanded="true" height="103" name="Split Data" width="90" x="313" y="289">
<enumeration key="partitions">
<parameter key="ratio" value="0.7"/>
<parameter key="ratio" value="0.3"/>
</enumeration>
<parameter key="sampling_type" value="stratified sampling"/>
</operator>
<operator activated="true" class="multiply" compatibility="7.0.001" expanded="true" height="103" name="Multiply" width="90" x="45" y="442"/>
<operator activated="true" class="filter_examples" compatibility="7.0.001" expanded="true" height="103" name="Filter Examples (6)" width="90" x="313" y="646">
<list key="filters_list">
<parameter key="filters_entry_key" value="Obstrucoes.equals.sim"/>
</list>
</operator>
<operator activated="true" class="filter_examples" compatibility="7.0.001" expanded="true" height="103" name="Filter Examples (5)" width="90" x="313" y="442">
<list key="filters_list">
<parameter key="filters_entry_key" value="Obstrucoes.equals.nao"/>
</list>
</operator>
<operator activated="true" class="sample" compatibility="7.0.001" expanded="true" height="82" name="Sample" width="90" x="514" y="442">
<parameter key="sample_size" value="1800"/>
<list key="sample_size_per_class"/>
<list key="sample_ratio_per_class"/>
<list key="sample_probability_per_class"/>
</operator>
<operator activated="true" class="append" compatibility="7.0.001" expanded="true" height="103" name="Append" width="90" x="648" y="544"/>
<operator activated="true" class="parallel_decision_tree" compatibility="7.0.001" expanded="true" height="82" name="Decision Tree" width="90" x="849" y="544">
<parameter key="criterion" value="gini_index"/>
<parameter key="maximal_depth" value="8"/>
<parameter key="minimal_gain" value="0.001"/>
</operator>
<operator activated="true" class="apply_model" compatibility="7.0.001" expanded="true" height="82" name="Apply Model" width="90" x="983" y="238">
<list key="application_parameters"/>
</operator>
<operator activated="true" class="performance" compatibility="7.0.001" expanded="true" height="82" name="Performance" width="90" x="1050" y="544"/>
<connect from_op="Retrieve Desentupimentos" from_port="output" to_op="Set Role" to_port="example set input"/>
<connect from_op="Set Role" from_port="example set output" to_op="Select Attributes" to_port="example set input"/>
<connect from_op="Select Attributes" from_port="example set output" to_op="Filter Examples" to_port="example set input"/>
<connect from_op="Filter Examples" from_port="example set output" to_op="Filter Examples (2)" to_port="example set input"/>
<connect from_op="Filter Examples (2)" from_port="example set output" to_op="Filter Examples (3)" to_port="example set input"/>
<connect from_op="Filter Examples (3)" from_port="example set output" to_op="Filter Examples (4)" to_port="example set input"/>
<connect from_op="Filter Examples (4)" from_port="example set output" to_op="Split Data" to_port="example set"/>
<connect from_op="Split Data" from_port="partition 1" to_op="Multiply" to_port="input"/>
<connect from_op="Split Data" from_port="partition 2" to_op="Apply Model" to_port="unlabelled data"/>
<connect from_op="Multiply" from_port="output 1" to_op="Filter Examples (5)" to_port="example set input"/>
<connect from_op="Multiply" from_port="output 2" to_op="Filter Examples (6)" to_port="example set input"/>
<connect from_op="Filter Examples (6)" from_port="example set output" to_op="Append" to_port="example set 2"/>
<connect from_op="Filter Examples (5)" from_port="example set output" to_op="Sample" to_port="example set input"/>
<connect from_op="Sample" from_port="example set output" to_op="Append" to_port="example set 1"/>
<connect from_op="Append" from_port="merged set" to_op="Decision Tree" to_port="training set"/>
<connect from_op="Decision Tree" from_port="model" to_op="Apply Model" to_port="model"/>
<connect from_op="Apply Model" from_port="labelled data" to_op="Performance" to_port="labelled data"/>
<connect from_op="Apply Model" from_port="model" to_port="result 3"/>
<connect from_op="Performance" from_port="performance" to_port="result 1"/>
<connect from_op="Performance" from_port="example set" to_port="result 2"/>
<portSpacing port="source_input 1" spacing="0"/>
<portSpacing port="sink_result 1" spacing="0"/>
<portSpacing port="sink_result 2" spacing="0"/>
<portSpacing port="sink_result 3" spacing="0"/>
<portSpacing port="sink_result 4" spacing="0"/>
<background height="219" location="//Samples/Tutorials/Basics/06/tutorial6" width="2000" x="12" y="12"/>
</process>
</operator>
</process>
0 -
Hi Paulo,
it is cool to see, that another civil engineer is working with RM. One of our Sales Engineers is actually civil engineer as well. Thomas Ott aka neuralmarkettrends (on twitter or youtube).
I think that this problem is indeed a good use case for data mining. I think that the usual points to look at are algorithm, Feature Selection and Feature Generation. Of course you can post data here and we as the community have a look on it.
~Martin0
Altair Employee