Regression via classification

韦塞尔

Hi,

I wish to run a classification algorithm on a regression task.
Generated a new attribute where the label attribute is now discrete.
Unfortunately, I can no longer, straightforward apply the Performance (Regression) operator.

I have to compute the absolute error myself using generate attributes.
Since I'm applying attribute selection, I'm doing this over and over again, which is really slow.
Is there a faster way to achieve this result?

Best regards,

Wessel

Attached an example process below

<帕拉meter key="use_exact_number" value="false"/>
<帕拉meter key="exact_number_of_attributes" value="-1"/>
<帕拉meter key="min_number_of_attributes" value="1"/>
<帕拉meter key="limit_max_number" value="true"/>
<帕拉meter key="max_number_of_attributes" value="2"/>


<帕拉meter key="leave_one_out" value="false"/>
<帕拉meter key="number_of_validations" value="10"/>
<帕拉meter key="sampling_type" value="stratified sampling"/>
<帕拉meter key="use_local_random_seed" value="false"/>
<帕拉meter key="local_random_seed" value="1992"/>


<帕拉meter key="k" value="3"/>
<帕拉meter key="weighted_vote" value="false"/>
<帕拉meter key="measure_types" value="MixedMeasures"/>
<帕拉meter key="mixed_measure" value="MixedEuclideanDistance"/>
<帕拉meter key="nominal_measure" value="NominalDistance"/>
<帕拉meter key="numerical_measure" value="EuclideanDistance"/>
<帕拉meter key="divergence" value="GeneralizedIDivergence"/>
<帕拉meter key="kernel_type" value="radial"/>
<帕拉meter key="kernel_gamma" value="1.0"/>
<帕拉meter key="kernel_sigma1" value="1.0"/>
<帕拉meter key="kernel_sigma2" value="0.0"/>
<帕拉meter key="kernel_sigma3" value="2.0"/>
<参数键= " kernel_degree " value = " 3.0 " / >
<帕拉meter key="kernel_shift" value="1.0"/>
<帕拉meter key="kernel_a" value="1.0"/>
<帕拉meter key="kernel_b" value="0.0"/>










<帕拉meter key="create_view" value="false"/>











<帕拉meter key="main_criterion" value="first"/>
<帕拉meter key="accuracy" value="true"/>
<帕拉meter key="classification_error" value="false"/>
<帕拉meter key="kappa" value="false"/>
<帕拉meter key="weighted_mean_recall" value="false"/>
<帕拉meter key="weighted_mean_precision" value="false"/>
<帕拉meter key="spearman_rho" value="false"/>
<帕拉meter key="kendall_tau" value="false"/>
<帕拉meter key="absolute_error" value="false"/>
<帕拉meter key="relative_error" value="false"/>
<帕拉meter key="relative_error_lenient" value="false"/>
<帕拉meter key="relative_error_strict" value="false"/>
<帕拉meter key="normalized_absolute_error" value="false"/>
<帕拉meter key="root_mean_squared_error" value="false"/>
<帕拉meter key="root_relative_squared_error" value="false"/>
<帕拉meter key="squared_error" value="false"/>
<帕拉meter key="correlation" value="false"/>
<帕拉meter key="squared_correlation" value="false"/>
<帕拉meter key="cross-entropy" value="false"/>
<帕拉meter key="margin" value="false"/>
<帕拉meter key="soft_margin_loss" value="false"/>
<帕拉meter key="logistic_loss" value="false"/>
<帕拉meter key="skip_undefined_labels" value="true"/>
<帕拉meter key="use_example_weights" value="true"/>




<帕拉meter key="prediction(leadTime)" value="parse([prediction(leadTime)])"/>
<帕拉meter key="leadTime" value="[leadTime_numeric]"/>

<帕拉meter key="keep_all" value="true"/>


<帕拉meter key="main_criterion" value="first"/>
<帕拉meter key="root_mean_squared_error" value="false"/>
<帕拉meter key="absolute_error" value="true"/>
<帕拉meter key="relative_error" value="false"/>
<帕拉meter key="relative_error_lenient" value="false"/>
<帕拉meter key="relative_error_strict" value="false"/>
<帕拉meter key="normalized_absolute_error" value="false"/>
<帕拉meter key="root_relative_squared_error" value="false"/>
<帕拉meter key="squared_error" value="false"/>
<帕拉meter key="correlation" value="false"/>
<帕拉meter key="squared_correlation" value="false"/>
<帕拉meter key="prediction_average" value="false"/>
<帕拉meter key="spearman_rho" value="false"/>
<帕拉meter key="kendall_tau" value="false"/>
<帕拉meter key="skip_undefined_labels" value="true"/>
<帕拉meter key="use_example_weights" value="true"/>



<帕拉meter key="rp" value="operator.RP.value.performance"/>
<帕拉meter key="cp" value="operator.CP.value.performance"/>
<帕拉meter key="fn" value="operator.Loop Subsets.value.feature_names"/>

<帕拉meter key="sorting_type" value="none"/>
<帕拉meter key="sorting_k" value="100"/>
<帕拉meter key="persistent" value="false"/>



<连接from_op = from_ CPport="example set" to_op="Generate Attributes" to_port="example set input"/>

MartinLiebig

Hi wessel,

i can't get your processes in - somethings wrong with the xml.

Can't you simply use parse numbers on the prediction, swap the the label and prediction roles on the right hand side of x-val and use standard performance operator?

~Martin