[
https://issues.apache.org/jira/browse/SPARK-15092?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Nick Pentreath updated SPARK-15092:
-----------------------------------
Assignee: holdenk
> toDebugString missing from ML DecisionTreeClassifier
> ----------------------------------------------------
>
> Key: SPARK-15092
> URL: https://issues.apache.org/jira/browse/SPARK-15092
> Project: Spark
> Issue Type: Bug
> Components: ML
> Affects Versions: 1.6.0
> Environment: HDP 2.3.4, Red Hat 6.7
> Reporter: Ivan SPM
> Assignee: holdenk
> Priority: Minor
> Labels: features
>
> The attribute toDebugString is missing from the DecisionTreeClassifier and
> DecisionTreeClassifierModel from ML. The attribute exists on the MLLib
> DecisionTree model.
> There's no way to check or print the model tree structure from the ML.
> The basic code for it is this:
> rom pyspark.ml import Pipeline
> from pyspark.ml.feature import VectorAssembler, StringIndexer
> from pyspark.ml.classification import DecisionTreeClassifier
> cl = DecisionTreeClassifier(labelCol='target_idx', featuresCol='features')
> pipe = Pipeline(stages=[target_index, assembler, cl])
> model = pipe.fit(df_train)
> # Prediction and model evaluation
> predictions = model.transform(df_test)
> mc_evaluator = MulticlassClassificationEvaluator(
> labelCol="target_idx", predictionCol="prediction", metricName="precision" )
> accuracy = mc_evaluator.evaluate(predictions)
> print("Test Error = {}".format(1.0 - accuracy))
> now it would be great to be able to do what is being done on the MLLib model:
> print model.toDebugString(), # it already has newline
> DecisionTreeModel classifier of depth 1 with 3 nodes
> If (feature 0 <= 0.0)
> Predict: 0.0
> Else (feature 0 > 0.0)
> Predict: 1.0
> but there's no toDebugString attribute either to the pipeline model or the
> DecisionTreeClassifier model:
> cl.toDebugString()
> Attribute Error
> https://spark.apache.org/docs/1.6.0/api/python/_modules/pyspark/mllib/tree.html
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
