water
Class OldModel

java.lang.Object
  water.Iced
      water.OldModel

All Implemented Interfaces:

java.lang.Cloneable, Freezable

Direct Known Subclasses:

DGLM.GLMModel, DPCA.PCAModel, KMeansModel, RFModel

public abstract class OldModel
extends Iced

A Model models reality (hopefully). A model can be used to 'score' a row, or a collection of rows on any compatible dataset - meaning the row has all the columns with the same names as used to build the mode.

Field Summary

Key	_dataKey Dataset key used to *build* the model, for models for which this makes sense, or null otherwise.
Key	_selfKey Key associated with this OldModel, if any.
ValueArray	_va Columns used in the model.
static DocGen.FieldDoc[]	DOC_FIELDS

Constructor Summary

OldModel()
Empty constructor for deserialization

OldModel(Key key)

OldModel(Key key, int[] cols, Key dataKey)
Default model, built from the selected columns of the given dataset.

OldModel(Key key, OldModel m)
Simple shallow copy constructor

OldModel(Key key, java.lang.String[] colNames, java.lang.String[] classNames)
Default artificial model, built from given column names.

OldModel(Key key, ValueArray va, Key dataKey)
Artificial model.

Method Summary

OldModel	adapt(java.lang.String[] colNames) Adapt model for given columns.
OldModel	adapt(ValueArray ary) Adapt model for the given dataset.
boolean	columnFilter(ValueArray.Column C)
int[]	columnMapping(java.lang.String[] names) Map from the model's columns to the given column names, or to -1 if no column name maps.
void	delete() Called when deleting this model, to cleanup any internal keys
void	fromJson(com.google.gson.JsonObject json)
static boolean	identityMap(int[] mapping) Check that this is the identity map
static boolean	isCompatible(int[] mapping) Check if this mapping is compatible.
boolean	isCompatible(java.lang.String[] names) Check if this model is compatible with this collection of column names.
boolean	isCompatible(ValueArray data) Check if this dataset is compatible with this model.
ValueArray.Column	response() Response column info
java.lang.String	responseName()
double	score(double[] data)
double	score(ValueArray ary, AutoBuffer bits, int rid)
protected abstract double	score0(double[] data)
protected double	score0(ValueArray data, AutoBuffer ab, int row_in_chunk) Bulk scoring API, on a compatible ValueArray (when pushed throw the mapping)
protected double	score0(ValueArray data, int row) Single row scoring, on a compatible ValueArray (when pushed throw the mapping)
com.google.gson.JsonObject	toJson()

Methods inherited from class water.Iced

clone, frozenType, init, newInstance, read, toDocField, write, writeJSON, writeJSONFields

Methods inherited from class java.lang.Object

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

Field Detail

DOC_FIELDS

public static DocGen.FieldDoc[] DOC_FIELDS

_selfKey

public final Key _selfKey

Key associated with this OldModel, if any.

_va

public final ValueArray _va

Columns used in the model. No dataset needs to be mapped to this ValueArray, it is just used to control for valid column data. The mean & sigma are from the training dataset listed below, and are used when normalizing scoring data. The Column names are used to match up with scoring data columns. The last Column is the response column.

_dataKey

public final Key _dataKey

Dataset key used to *build* the model, for models for which this makes sense, or null otherwise. Not all models are built from a dataset (eg artificial models), or are built from a single dataset (various ensemble models), so this key has no *mathematical* significance in the model but is handy during common model-building and for the historical record.

Constructor Detail

OldModel

public OldModel()

Empty constructor for deserialization

OldModel

public OldModel(Key key)

OldModel

public OldModel(Key key,
                int[] cols,
                Key dataKey)

Default model, built from the selected columns of the given dataset. Data to be scored on the model has to have all the same columns (in any order, extra cols are ok). Last column is the response column, or -1 if there is no defined response column.

OldModel

public OldModel(Key key,
                java.lang.String[] colNames,
                java.lang.String[] classNames)

Default artificial model, built from given column names.

OldModel

public OldModel(Key key,
                ValueArray va,
                Key dataKey)

Artificial model. The 'va' defines the compatible data, but is not associated with any real dataset. Data to be scored on the model has to have all the same columns (in any order, extra cols are ok). The last column is the response column.

OldModel

public OldModel(Key key,
                OldModel m)

Simple shallow copy constructor

Method Detail

columnFilter

public boolean columnFilter(ValueArray.Column C)

delete

public void delete()

Called when deleting this model, to cleanup any internal keys

response

public final ValueArray.Column response()

Response column info

responseName

public final java.lang.String responseName()

columnMapping

public final int[] columnMapping(java.lang.String[] names)

Map from the model's columns to the given column names, or to -1 if no column name maps. Last entry is a mapping for the response Name. Return results range from -1 to number-of-columns in the dataset/names[] (which may be larger than the model).

identityMap

public static boolean identityMap(int[] mapping)

Check that this is the identity map

isCompatible

public static boolean isCompatible(int[] mapping)

Check if this mapping is compatible. Just means no -1 entries in the predictor variables (response column is not checked).

isCompatible

public final boolean isCompatible(java.lang.String[] names)

Check if this model is compatible with this collection of column names.

isCompatible

public final boolean isCompatible(ValueArray data)

Check if this dataset is compatible with this model. All the columns in the model have to be present, but extra columns may exist and the columns may be in a different order.

adapt

public OldModel adapt(ValueArray ary)

Adapt model for the given dataset. Default behavior is to map columns and categoricals to their original indexes. Categorical values we have not seen when building the model are translated as NaN. Override this to get custom adapt behavior (eg. handle unseen cats differently).

Parameters:

ary - - tst dataset

Returns:

OldModel - model adapted to be applied on the given data

adapt

public OldModel adapt(java.lang.String[] colNames)

Adapt model for given columns. Only permutes the columns by the column names (factor levels MUST match the training dataset).

Parameters:

colNames -

Returns:

score

public double score(double[] data)

score

public double score(ValueArray ary,
                    AutoBuffer bits,
                    int rid)

score0

protected abstract double score0(double[] data)

score0

protected double score0(ValueArray data,
                        int row)

Single row scoring, on a compatible ValueArray (when pushed throw the mapping)

score0

protected double score0(ValueArray data,
                        AutoBuffer ab,
                        int row_in_chunk)

Bulk scoring API, on a compatible ValueArray (when pushed throw the mapping)

toJson

public com.google.gson.JsonObject toJson()

fromJson

public void fromJson(com.google.gson.JsonObject json)