org.knime.base.node.mine.bayes.naivebayes.datamodel
Class AttributeModel

java.lang.Object
  org.knime.base.node.mine.bayes.naivebayes.datamodel.AttributeModel

All Implemented Interfaces:

Comparable<AttributeModel>

Direct Known Subclasses:

BitVectorAttributeModel, ClassAttributeModel, NominalAttributeModel, NumericalAttributeModel

public abstract class AttributeModel
extends Object
implements Comparable<AttributeModel>

This abstract class needs to be implemented by all attribute models and provides missing value handling and some common methods.

Author:

Tobias Koetter, University of Konstanz

Field Summary

(package private) static String	MODEL_CONTAINS_NO_CLASS_VALUES Invalid cause if the model contains no class values.
(package private) static String	MODEL_CONTAINS_NO_RECORDS Invalid cause if the model contains no records at all.

Constructor Summary

AttributeModel(String attributeName, int noOfMissingVals, boolean skipMissingVals)
Constructor for class ClassValue.

Method Summary

(package private) void	addValue(String classValue, DataCell attrValue) Adds the given value to this attribute model.
(package private) abstract void	addValueInternal(String classValue, DataCell attrValue) Adds the given value to the concrete implementation.
int	compareTo(AttributeModel o)
(package private) static String	createHTMLTable(String tableHeading, String keyHeading, String valueHeading, int noOfRows, Map<String,? extends Object> map, boolean addLineBreak)
protected static String	createTableHeader(String firstHeading, List<String> headings, String lastHeading)
String	getAttributeName()
(package private) abstract Collection<String>	getClassValues()
(package private) abstract Class<? extends DataValue>	getCompatibleType()
(package private) abstract String	getHTMLView(int totalNoOfRecs)
(package private) abstract String	getHTMLViewHeadLine()
String	getInvalidCause()
(package private) String	getMissingValueHeader(Collection<String> colNames)
(package private) int	getNoOfMissingVals()
(package private) abstract Integer	getNoOfRecs4ClassValue(String classValue)
(package private) Double	getProbability(String classValue, DataCell attributeValue, double laplaceCorrector)
(package private) abstract double	getProbabilityInternal(String classValue, DataCell attributeValue, double laplaceCorrector) This should also handle missing values.
(package private) abstract String	getType()
(package private) static AttributeModel	loadModel(Config config)
(package private) void	saveModel(Config config)
(package private) abstract void	saveModelInternal(Config config)
(package private) void	setInvalidCause(String cause)
(package private) static List<String>	sortCollection(Collection<String> vals)
(package private) abstract void	validate() Called after all training rows where added to validate the model.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

MODEL_CONTAINS_NO_RECORDS

static final String MODEL_CONTAINS_NO_RECORDS

Invalid cause if the model contains no records at all.

See Also:

Constant Field Values

MODEL_CONTAINS_NO_CLASS_VALUES

static final String MODEL_CONTAINS_NO_CLASS_VALUES

Invalid cause if the model contains no class values.

See Also:

Constant Field Values

Constructor Detail

AttributeModel

AttributeModel(String attributeName,
               int noOfMissingVals,
               boolean skipMissingVals)

Constructor for class ClassValue.

Parameters:

attributeName - the name of the attribute

noOfMissingVals - the number of missing values

skipMissingVals - set to true if the missing values should be skipped during learning and prediction

Method Detail

loadModel

static AttributeModel loadModel(Config config)
                         throws org.knime.core.node.InvalidSettingsException

Parameters:

config - the Config object to read from

Returns:

the attribute model for the given Config object

Throws:

org.knime.core.node.InvalidSettingsException - if the settings are invalid

saveModel

void saveModel(Config config)

Parameters:

config - the Config object to write to

saveModelInternal

abstract void saveModelInternal(Config config)

Parameters:

config - the config object to save to

getType

abstract String getType()

Returns:

the unique type of the model

getAttributeName

public String getAttributeName()

Returns:

the attribute name

getCompatibleType

abstract Class<? extends DataValue> getCompatibleType()

Returns:

the DataValue class to check if the rows are compatible

addValue

void addValue(String classValue,
              DataCell attrValue)
        throws TooManyValuesException

Adds the given value to this attribute model.

Parameters:

classValue - the class value

attrValue - the attribute value. Could be a missing value.

Throws:

TooManyValuesException - if the column contains more unique values than supported by this attribute model

addValueInternal

abstract void addValueInternal(String classValue,
                               DataCell attrValue)
                        throws TooManyValuesException

Adds the given value to the concrete implementation. Should handle missing values as well.

Parameters:

classValue - the class value

attrValue - the attribute value. Could be a missing value.

Throws:

TooManyValuesException - if the column contains more unique values than supported by this attribute model

getNoOfMissingVals

int getNoOfMissingVals()

Returns:

the noOfMissingVals

getMissingValueHeader

String getMissingValueHeader(Collection<String> colNames)

Parameters:

colNames - all column names of the table to check for uniqueness

Returns:

the missing value header or null if this model contains no missing attribute values

getClassValues

abstract Collection<String> getClassValues()

Returns:

all class values

getNoOfRecs4ClassValue

abstract Integer getNoOfRecs4ClassValue(String classValue)

Parameters:

classValue - the class value we want the number of records for

Returns:

the number of records with the given class value or null if only missing values where in this row for the given class value or the class value wasn't found at all

getProbability

Double getProbability(String classValue,
                      DataCell attributeValue,
                      double laplaceCorrector)

Parameters:

classValue - the class value to calculate the probability for

attributeValue - the attribute value to calculate the probability for. Could be a missing value.

laplaceCorrector - the Laplace corrector to use. A value greater 0 overcomes zero counts.

Returns:

the calculated probability or null if the cell was a missing one and missing values should be skipped

getProbabilityInternal

abstract double getProbabilityInternal(String classValue,
                                       DataCell attributeValue,
                                       double laplaceCorrector)

This should also handle missing values.

Parameters:

classValue - the class value to calculate the probability for

attributeValue - the attribute value to calculate the probability for. Could be a missing value.

laplaceCorrector - the Laplace corrector to use. A value greater 0 overcomes zero counts.

Returns:

the calculated probability

getHTMLView

abstract String getHTMLView(int totalNoOfRecs)

Parameters:

totalNoOfRecs - the total number of records in the training data

Returns:

the HTML view of this attribute model

sortCollection

static List<String> sortCollection(Collection<String> vals)

Parameters:

vals - the Collection to sort

Returns:

the given Collection in her natural order

validate

abstract void validate()
                throws org.knime.core.node.InvalidSettingsException

Called after all training rows where added to validate the model.

Throws:

org.knime.core.node.InvalidSettingsException - if the model isn't valid

setInvalidCause

void setInvalidCause(String cause)

Parameters:

cause - the cause why this model is invalid

getInvalidCause

public String getInvalidCause()

Returns:

if the model is invalid this method returns the reason why otherwise it returns null.

compareTo

public int compareTo(AttributeModel o)

Specified by:

compareTo in interface Comparable<AttributeModel>

getHTMLViewHeadLine

abstract String getHTMLViewHeadLine()

Returns:

the headline of this model to use in the HTML view

createHTMLTable

static String createHTMLTable(String tableHeading,
                              String keyHeading,
                              String valueHeading,
                              int noOfRows,
                              Map<String,? extends Object> map,
                              boolean addLineBreak)

Parameters:

tableHeading - the optional table headline

keyHeading - the optional headline for the key row

valueHeading - the optional headline for the value row

noOfRows - the number of rows displayed per row

map - the map to create the html table for

addLineBreak - if each sub table should be displayed on a new line

Returns:

a html table with the keys and values of the given map

createTableHeader

protected static String createTableHeader(String firstHeading,
                                          List<String> headings,
                                          String lastHeading)

Parameters:

firstHeading - the optional first heading

headings - the head lines

lastHeading - the optional last heading

Returns:

the string with a html table head line row