PREDICT_RF_CLASSIFIER

Applies a random forest model on an input table or view. The predicted class is selected only based on the popular vote of the decision trees in the forest. Therefore, in special cases the calculated probability of the predicted class may not be the highest.

Important: Before using a machine learning function, be aware that all the ongoing transactions might be committed.

Syntax

PREDICT_RF_CLASSIFIER ( col1, col2, ... coln
                       USING PARAMETERS model_name = 'name_of_model'
                                                    [, type= ' { RESPONSE | PROBABILITY } ',]                  
                                                    [class= 'user_input_class',]
                                                    [match_by_pos = 'method'])    
                                

Arguments

col1, col2, ..., coln

The columns to use from the input table or view.

Parameters

model_name = 'name_of_model'

The name of the random forest model. Model names are case-insensitive.

type = 'method'

(Optional) Determines the type of prediction for random forest.

When response is selected the result of the prediction is the class with the highest probability among all possible classes. When probability is selected, the 'class' parameter is considered.

Valid Values

  • response (Default Value)
  • probability
class = 'user_input_class'

(Optional) Specifies a specific class to use when type is set to PROBABILITY. The predict function returns the probability of the specified class to be the response. If class is not specified, its default value is the predicted class -- the one with popular vote. Thus, the predict function returns the probability that the input instance belonging to its predicted class.

Default Value: Auto

match_by_pos= 'method'

(Optional) Valid Values:

  • false (default): Input columns will be matched to features in the model based on their names.

  • true: Input columns will be matched to features in the model based on their position in the list of indicated input columns.

Return

Return data type: VARCHAR

The predict function returns the predicted class (based on popular votes) or probability of a class (depending on the value of the optional input parameters type and class) for each input instance.

Examples

This example shows how you can use the PREDICT_RF_CLASSIFIER function.

=> SELECT PREDICT_RF_CLASSIFIER (Sepal_Length, Sepal_Width, Petal_Length, Petal_Width
                                  USING PARAMETERS model_name='myRFModel') FROM iris;
PREDICT_RF_CLASSIFIER
-----------------------
setosa
setosa
setosa
.
.
.
versicolor
versicolor
versicolor
.
.
.
virginica
virginica
virginica
.
.
.
(150 rows)

This example shows how you can use the PREDICT_RF_CLASSIFIER function, using the match_by_pos parameter:

=> SELECT PREDICT_RF_CLASSIFIER (Sepal_Length, Sepal_Width, Petal_Length, Petal_Wdith
                                USING PARAMETERS model_name='myRFModel', match_by_pos='true') FROM iris;
PREDICT_RF_CLASSIFIER
-----------------------
setosa
setosa
setosa
.
.
.
versicolor
versicolor
versicolor
.
.
.
virginica
virginica
virginica
.
.
.
(150 rows)

 

See Also