PMML 4.2 - Changes from PMML 4.1
PMML4.2 Menu

Home

PMML Notice and License

Changes

XML Schema

Conformance

Interoperability

General Structure

Field Scope

Header

Data
Dictionary


Mining
Schema


Transformations

Statistics

Taxomony

Targets

Output

Functions

Built-in Functions

Model Verification

Model Explanation

Multiple Models

Association Rules

Baseline Models

Cluster
Models


General
Regression


k-Nearest
Neighbors


Naive
Bayes


Neural
Network


Regression

Ruleset

Scorecard

Sequences

Text Models

Time Series

Trees

Vector Machine

PMML 4.2.1 - Changes from PMML 4.2


Transformations

  • Clarified the usage of attribute 'wordCharacterSeparatorRE' in the TextIndex element
  • Added attribute 'tokenize' to the TextIndex element and explanation of its usage
  • Corrected example with the TextIndexNormalization element


PMML 4.2 - Changes from PMML 4.1


Associations

  • Changed version to "4.2" from "4.1" in the example(s)

BaselineModel

  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element MiningSchema from "predicted" to "target" in the example(s)

BuiltinFunctions

  • Added built-in functions matches, replace, and concat

ClusteringModel

  • Added text explaining the usage of attribute 'numberOfClusters'
  • Updated minimun number of occurrences of element ClusteringField to "1"
  • Changed version to "4.2" from "4.1" in the example(s)

Conformance

  • Changed all version references from "4.1" to "4.2"

DataDictionary

  • No changes

FieldScope

  • No changes

Functions

  • Added 'defaultValue' attribute to the Apply element
  • Clarified the differences between the use of attributes 'mapMissingTo' and 'defaultValue' attributes for the Apply element and added a table outlining all the possible outputs for the Apply element

GeneralRegression

  • Clarified the order of ordinal fields in the text and updated the ordinal multinomial example accordingly
  • Corrected attribute 'modelType' from 'generalLinear' to 'generalizedLinear' for the generalized linear example
  • Added attribute 'targetReferenceCategory' and list of valid values in element 'DataDictionary' for the multinomial logistic example
  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element 'MiningSchema' from "predicted" to "target" in the example(s)

GeneralStructure

  • Updated the XML namespace to "https://www.dmg.org/PMML-4_2" from "https://www.dmg.org/PMML-4_1"

Header

  • No changes

Interoperability

  • No changes

KNN

  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element 'MiningSchema' from "predicted" to "target" in the example(s)

MiningSchema

  • Added value "target" for attribute 'usageType'. Deprecated value "predicted"

ModelExplanation

  • No changes

ModelVerification

  • No changes

MultipleModels

  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element 'MiningSchema' from "predicted" to "target" in the example(s)

NaïveBayes

  • Modified introductory text and table to show how Naïve Bayes models handle continuous input fields
  • Added element TargetValueStats for the handling of continuous input fields
  • Modified example to show the usage of new element TargetValueStats through variable "age of individual". Updated scoring procedure accordingly
  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element 'MiningSchema' from "predicted" to "target" in the example(s)

NeuralNetwork

  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element MiningSchema from "predicted" to "target" in the example(s)

Output

  • Generalized the use of attribute 'rank' and made it more consistent
  • Revised and clarified the text for attribute 'segmentId'
  • Deprecated attribute 'ruleFeature'. Rule feature values are now specified directly in attribute 'feature'
  • Deprecated attribute 'isMultiValued'
  • Deprecated result features "ruleValue" and "ruleId"
  • Updated text for element Decisions to describe its usage together with transformed values in representing data post-processing
  • Updated table describing the output per model type
  • Added text specifying treatment of output values for a model chain
  • Updated text describing result feature "residual"

Regression

  • Corrected example for binary logistic regression
  • Clarified the usage of element Regression for implementing binary logistic regression models
  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element MiningSchema from "predicted" to "target" in the example(s)

RuleSet

  • Added a default value of "1" for attributes 'confidence' and 'weight'
  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element MiningSchema from "predicted" to "target" in the example(s)

Scorecard

  • Added element ComplexPartialScore and updated example to show its usage
  • Added text to define the how reason codes should be output in case of a tie
  • Added text to define what should be returned in case not even a single attribute/predicate evaluates to true for a given characteristic
  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element MiningSchema from "predicted" to "target" in the example(s)

Sequences

  • Changed version to "4.2" from "4.1" in the example(s)

Statistics

  • No changes

SupportVectorMachine

  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element MiningSchema from "predicted" to "target" in the example(s)

Targets

  • No changes

Taxonomy

  • Added element FieldColumnPair in ChildParent element

TextModel

  • Deprecated element TextModel

TimeSeriesModel

  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element MiningSchema from "predicted" to "target" in the example(s)

Transformations

  • Clarified the usage of attribute 'mapMissingTo' in the FieldRef element
  • Added minOccurs="0" in element MapValues for FieldColumnPair
  • Added new element TextIndex for extracting term frequencies from text

TreeModel

  • Changed version to "4.2" from "4.1" in the example(s)
  • Changed value of attribute 'usageType' in element MiningSchema from "predicted" to "target" in the example(s)
e-mail info at dmg.org