Skip navigation links

oracle.dmt.jdm.algorithm.glm
Interface OraGLMSettings

All Superinterfaces:

javax.datamining.base.AlgorithmSettings, javax.datamining.supervised.SupervisedAlgorithmSettings

All Known Subinterfaces:

OraGLMClassificationSettings, OraGLMRegressionSettings

public interface OraGLMSettings

extends javax.datamining.supervised.SupervisedAlgorithmSettings

OraGLMSettings is the super interface for GLM classification and regression algorithm settings interface. It defines the common GLM setting, that includes:

• Ridge regression - Enable ridge regression to build a better model when the input data has correlated attributes (multi-collinearity). By default ridge regression enabling flag is set to null, so that system determines it.
• Diagnostics Table Name - GLM can produce per case/row diagnostics as part of GLM model building. User can optionally specify diagnostics table name to output per row diagnostics that can be used to validate the model. These diagnostics include various types of residuals, influence measures, and the training data predictions. By default no diagnostics table will be created, user has to explicitely specify the diagnostics table name to create one as a by-product of model build.

  For classification models, the diagnostics table will have the following schema:

  1. CASE_ID - Unique identifier for each row of the model build input table
  2. TARGET_VALUE - A value of the target attribute
  3. TARGET_VALUE_PROB - Probability associated with the target value
  4. HAT - Diagonal value of the hat matrix
  5. WORKING_RESIDUAL - Working residuals
  6. PEARSON_RESIDUAL - Pearson residuals
  7. DEVIANCE_RESIDUAL - Deviance residuals
  8. C - Confidence interval displacement diagnostic
  9. CBAR - Confidence interval displacement diagnostic

  For regression model, diagnostics table will have the following schema:

  1. CASE_ID - Unique identifier for each row of the model build input table
  2. TARGET_VALUE - Actual value of the target
  3. PREDICTED_VALUE - Predicted value of the target
  4. HAT - Diagonal value of the hat matrix
  5. RESIDUAL - Normal residual
  6. STD_ERR_RESIDUAL - Standardized residual
  7. STUDENTIZED_RESIDUAL - Studentized residual
  8. PRED_RES -
  9. COOKS_D - Cooks distance

Since:

OJDM 11.1.0

Field Summary

static java.lang.String	TREAT_MISSING_VALUE_AS_CASE_EXCLUSION Constant used to specify missing values are treated by excluding the whole case when any of the model singature attribute values are missing.
static java.lang.String	TREAT_MISSING_VALUE_AS_MEAN_OR_MODE Constant used to specify missing values are treated using the mean value for numerical attributes and mode value for categorical attributes.

Method Summary

double	getConfidenceLevel() Returns the specified confidence level for coefficient bounds.
java.lang.String	getDiagnosticsTableName() Returns the specified diagnostics table name.
java.lang.String	getMissingValueTreatmentType() Returns the missing value treatment type.
java.lang.Double	getRidgeValue() Returns the user specified ridge value.
void	setConfidenceLevel(double confidenceLevel) Sets confidence level for coefficient bounds.
void	setDiagnosticsTableName(java.lang.String diagnosticsTableName) Sets the optional row diagnostics table name.
void	setMissingValueTreatmentType(java.lang.String treatmentType) Set the missing value treatment type.
void	setRidgeValue(java.lang.Double ridgeVal) Sets the user specified ridge value.
java.lang.Boolean	useRidgeRegression() Returns true when ridge regression is enabled.
void	useRidgeRegression(java.lang.Boolean enableRidgeRegression) Enables ridge regression when it is set to true.

Methods inherited from interface javax.datamining.base.AlgorithmSettings

getMiningAlgorithm, verify

Field Detail

TREAT_MISSING_VALUE_AS_MEAN_OR_MODE

public static final java.lang.String TREAT_MISSING_VALUE_AS_MEAN_OR_MODE

Constant used to specify missing values are treated using the mean value for numerical attributes and mode value for categorical attributes.

See Also:

Constant Field Values

TREAT_MISSING_VALUE_AS_CASE_EXCLUSION

public static final java.lang.String TREAT_MISSING_VALUE_AS_CASE_EXCLUSION

Constant used to specify missing values are treated by excluding the whole case when any of the model singature attribute values are missing.

See Also:

Constant Field Values

Method Detail

useRidgeRegression

public java.lang.Boolean useRidgeRegression()

Returns true when ridge regression is enabled. Otherwise returns false. Returns null when the system determines the ridge regression to be enabled or disabled.

Returns:

useRidgeRegression

public void useRidgeRegression(java.lang.Boolean enableRidgeRegression)

Enables ridge regression when it is set to true. Otherwise ridge regression is diabled. When specified as null system determines whether to enable/disable ridge regression.

Parameters:

enableRidgeRegression -

getRidgeValue

public java.lang.Double getRidgeValue()

Returns the user specified ridge value. Returns null when the ridge value is set to be determined by the algorithm implemenation.

setRidgeValue

public void setRidgeValue(java.lang.Double ridgeVal)

Sets the user specified ridge value. When specified as a not-null value then ridge regression is implicitely enabled. When specified as a null value then the algorithm implementation will determine the ridge value.

Parameters:

ridgeVal -

setConfidenceLevel

public void setConfidenceLevel(double confidenceLevel)

Sets confidence level for coefficient bounds. By default it is pre-set to 0.95.

Parameters:

confidenceLevel - valid value is greater than 0 and less than 1

getConfidenceLevel

public double getConfidenceLevel()

Returns the specified confidence level for coefficient bounds.

Returns:

setDiagnosticsTableName

public void setDiagnosticsTableName(java.lang.String diagnosticsTableName)

Sets the optional row diagnostics table name. When diagnostics table is not specified, algorithm do not create this table.

Parameters:

diagnosticsTableName -

getDiagnosticsTableName

public java.lang.String getDiagnosticsTableName()

Returns the specified diagnostics table name.

Returns:

setMissingValueTreatmentType

public void setMissingValueTreatmentType(java.lang.String treatmentType)

Set the missing value treatment type. There are two types of missing value treatments allowed:

• Use mean value for numerical attributes and mode value for categorical attributes.
• Exclude the whole case if any of the model singature attribute values miss in case.

By default mean or mode value settings are used.

getMissingValueTreatmentType

public java.lang.String getMissingValueTreatmentType()

Returns the missing value treatment type.

Returns:

Skip navigation links