| Previous | Table of Contents | Next |
A MiningModel holds the metadata of the result of a mining (training) task. This information is sufficient to determine whether
a model can be applied to a given data.
Superclasses
ModelElement
Contained Elements
ModelSignature
Attributes
function
Data mining function (as opposed to algorithm); for example, classification or clustering, attributeImportance, associationRules,
classification, approximation, clustering
type: MiningFunction (attributeImportance | associationRules | classification | approximation | clustering)
multiplicity: exactly one
algorithmName
Specific implementation of the data mining function (e.g., CART decision tree or SOM clustering). The following algorithm
names are predefined (their functions in parentheses):
• decisionTree (classification, approximation)
• neuralNetwork (classification, approximation)
• naiveBayes (classification)
• selfOrganizingMap (clustering)
• kMeans (clustering)
• competitiveLearning
type: String
multiplicity: exactly one
keyValue
This optionally represents the key value when the model is to be located.type: Anymultiplicity: exactly one
References
settings
The settings that were used to generate the model.class: MiningFunctionSettingsdefined by: ModelRefSettingsmultiplicity: zero or oneinverse: MiningFunctionSettings::model
modelSignature
The set of attributes (SignatureAttributes) that were used to build the model.class: ModelSignaturedefined by: ModelHasSignaturemultiplicity: zero or oneinverse: ModelSignature::modelaggregation: composite
modelLocation
This optionally provides a way to locate the model in the metadata repository. class: Class defined by: MiningModelRefLocation
multiplicity: exactly one inverse: Class::model
keyAttribute
This optionally identifies the key attribute when the model is located via modelLocation. class: Attribute defined by: ModelRefKeyAttribute
multiplicity: exactly one inverse: Attribute::model