[R] PMML for Cox Regression

[Rick Pechter]

Hi,

We have created a cox regression model (coxph) and exported from R as PMML
using the Rattle package.  The resulting PMML is a standard Regression model
in the format y=b1*x1+b2*x2+…….bn*xn:

risk= (0.027968680616809*dropped_Calls)+(0.00777220409115466*helpdesk_Calls)

The PMML is included at the end of this post.

Cox Regression is usually in this format:

h(t)= h0(t)*e^(b1*x1+b2*x2+…….bn*xn) 

We couldn't find any documentation about how a scoring engine should
generate scores from the R PMML and it seems that the standard regression
model will not generate the proper results for this cox regression model. 
Can someone point us to the proper documentation for scoring this model?

Thanks in advance, 

Rick Pechter

<?xml version="1.0"?>
<PMML version="3.2" xmlns="http://www.dmg.org/PMML-3_2"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://www.dmg.org/PMML-3_2
http://www.dmg.org/v3-2/pmml-3-2.xsd">
 <Header copyright="Copyright (c) 2011 sdoberman" description="CoxPH
Survival Regression Model">
  <Extension name="user" value="sdoberman" extender="Rattle/PMML"/>
  <Application name="Rattle/PMML" version="1.2.27"/>
  <Timestamp>2011-10-11 15:13:04</Timestamp>
 </Header>
 <DataDictionary numberOfFields="3">
  <DataField name="risk" optype="continuous" dataType="double"/>
  <DataField name="dropped_Calls" optype="continuous" dataType="double"/>
  <DataField name="helpdesk_Calls" optype="continuous" dataType="double"/>
 </DataDictionary>
 <RegressionModel modelName="CoxPH_Survival_Regression_Model"
functionName="regression" algorithmName="coxph" targetFieldName="risk">
  <MiningSchema>
   <MiningField name="risk" usageType="predicted"/>
   <MiningField name="dropped_Calls" usageType="active"/>
   <MiningField name="helpdesk_Calls" usageType="active"/>
  </MiningSchema>
  <RegressionTable intercept="0">
   <NumericPredictor name="dropped_Calls" exponent="1"
coefficient="0.027968680616809" mean="0.5919"/>
   <NumericPredictor name="helpdesk_Calls" exponent="1"
coefficient="0.00777220409115466" mean="0.5779"/>
  </RegressionTable>
 </RegressionModel>
</PMML>


--
View this message in context: http://r.789695.n4.nabble.com/PMML-for-Cox-Regression-tp3896791p3896791.html
Sent from the R help mailing list archive at Nabble.com.