"Estimate performance of SVM for regression"

User: "noah977"
noah977
New Altair Community Member
Updated by Jocelyn
Hi,

I am building some model files using SVM for REGRESSION (nu-SVR)

I can't seem to find a way to see a performance estimate from this task.  Either when training the SVM or using the model file.

I tried the "performance" operator, but it gives me errors.

Any ideas?

Find more posts tagged with

AI Studio
Performance
SVMs
Regression
Sort by:
1 - 2 of 21
    User: "earmijo"
    earmijo
    New Altair Community Member
    Go thru the following steps:

    1) Read your data
    2) Learn the model  (check the keep the example set)
    3) Apply the model
    4) Regression performance operator

    Here's a modification of the SupportVectorMachine.xml file that you can find in the samples (under Learners) 
    <operator name="Root" class="Process" expanded="yes">
        <description text="For many learning tasks, Support Vector Machines are among the best suited learning schemes.They adapt the idea of structural risk minimization and allows for non-linear generalizations with help of kernel functions."/>
        <operator name="ExampleSource" class="ExampleSource">
            <parameter key="attributes"	value="../data/polynomial.aml"/>
        </operator>
        <operator name="JMySVMLearner" class="JMySVMLearner">
            <parameter key="keep_example_set"	value="true"/>
            <parameter key="kernel_degree"	value="3"/>
            <parameter key="kernel_type"	value="polynomial"/>
        </operator>
        <operator name="ModelApplier" class="ModelApplier">
        </operator>
        <operator name="RegressionPerformance" class="RegressionPerformance">
            <parameter key="root_mean_squared_error"	value="true"/>
        </operator>
    </operator>
    User: "TobiasMalbrecht"
    TobiasMalbrecht
    New Altair Community Member
    Hi,

    please note that the solution earmijo gave calculates only the training error. To get a robust and general performance value you should use a cross validation instead.

    Regards,
    Tobias
Quick Links