MLlib Random Forest Classification Example with PySpark

          PySpark MLlib API provides a RandomForestClassifier class to classify data with random forest method. A random forest model is an ensemble learning algorithm based on decision tree learners. The model generates several decision trees and provides a combined result out of all outputs. Each tree in a forest votes and forest makes a decision based on all votes. A vote depends on the correlation between the trees and the strength of each tree.

    In this tutorial, we'll briefly learn how to train and classify data by using PySpark RandomForestClassifier. The tutorial covers:

1. Preparing the data
2. Prediction and accuracy check
3. Source code listing

   We'll start by loading the required libraries for this tutorial.

from pyspark import SparkContext
from pyspark.sql import SQLContext
from pyspark.ml.classification import RandomForestClassifier
from pyspark.ml.evaluation import MulticlassClassificationEvaluator
from pyspark.ml.feature import VectorAssembler
from sklearn.metrics import confusion_matrix
from sklearn.datasets import load_iris
import pandas as pd 

Preparing the data

   We use Iris dataset to perform classification and it can be easily loaded from the Scikit-learn dataset module. Below code explains how to load dataset and transform it into the pandas data frame type. 

iris = load_iris()
df_iris = pd.DataFrame(iris.data, columns=iris.feature_names)
df_iris['label'] = pd.Series(iris.target)

print(df_iris.head())

   sepal length (cm)  sepal width (cm)  ...  petal width (cm)  label
0                5.1               3.5  ...               0.2      0
1                4.9               3.0  ...               0.2      0
2                4.7               3.2  ...               0.2      0
3                4.6               3.1  ...               0.2      0
4                5.0               3.6  ...               0.2      0 

Next, we'll define SqlConext and create data frame by using df_iris data.

 

sc = SparkContext().getOrCreate()
sqlContext = SQLContext(sc)

data = sqlContext.createDataFrame(df_iris)
print(data.printSchema()) 

root
 |-- sepal length (cm): double (nullable = true)
 |-- sepal width (cm): double (nullable = true)
 |-- petal length (cm): double (nullable = true)
 |-- petal width (cm): double (nullable = true)
 |-- label: long (nullable = true)

 

To combine all feature data and separate 'label' data in a dataset, we use VectorAssembler.

features = iris.feature_names

va = VectorAssembler(inputCols = features, outputCol='features')

va_df = va.transform(data)
va_df = va_df.select(['features', 'label'])
va_df.show(3)
  

+-----------------+-----+
|         features|label|
+-----------------+-----+
|[5.1,3.5,1.4,0.2]|    0|
|[4.9,3.0,1.4,0.2]|    0|
|[4.7,3.2,1.3,0.2]|    0|
+-----------------+-----+
only showing top 3 rows

Next, we'll split data into the train and test parts.

(train, test) = va_df.randomSplit([0.8, 0.2])

 

Prediction and Accuracy Check

   Next, we'll define the decision tree classifier model by using the RandomForestClassifier class and fit model on train data. We can predict test data by using trasnform() method.

 

rfc = RandomForestClassifier(featuresCol="features", labelCol="label")
rfc = rfc.fit(train)

pred = rfc.transform(test)
pred.show(3) 

+-----------------+-----+--------------+-------------+----------+
|         features|label| rawPrediction|  probability|prediction|
+-----------------+-----+--------------+-------------+----------+
|[4.4,2.9,1.4,0.2]|    0|[20.0,0.0,0.0]|[1.0,0.0,0.0]|       0.0|
|[4.6,3.1,1.5,0.2]|    0|[20.0,0.0,0.0]|[1.0,0.0,0.0]|       0.0|
|[5.0,3.6,1.4,0.2]|    0|[20.0,0.0,0.0]|[1.0,0.0,0.0]|       0.0|
+-----------------+-----+--------------+-------------+----------+
only showing top 3 rows 

 

    After training the model, we'll predict test data and check the accuracy metrics. Here, we can use MulticlassClassificationEvaluator to check the accuracy. Confusion matrix can be created by using confusion_matrix function of sklearn.metrics module.

evaluator=MulticlassClassificationEvaluator(predictionCol="prediction")
acc = evaluator.evaluate(pred)
 

print("Prediction Accuracy: ", acc)

y_pred=pred.select("prediction").collect()
y_orig=pred.select("label").collect()

cm = confusion_matrix(y_orig, y_pred)
print("Confusion Matrix:")
print(cm)

Prediction Accuracy:  0.9674347158218126

Confusion Matrix:

[[13  0  0]
 [ 0 10  0]
 [ 0  1  7]] 

 

Finally, we'll stop spark context session.

 

# Stop session 
sc.stop()  

 

   In this tutorial, we've briefly learned how to fit and classify data by using PySpark RandomForestClassifier class. The full source code is listed below.

Source code listing

 

from pyspark import SparkContext
from pyspark.sql import SQLContext
from pyspark.ml.classification import RandomForestClassifier
from pyspark.ml.evaluation import MulticlassClassificationEvaluator
from pyspark.ml.feature import VectorAssembler
from sklearn.metrics import confusion_matrix
from sklearn.datasets import load_iris
import pandas as pd


iris = load_iris()
df_iris = pd.DataFrame(iris.data, columns=iris.feature_names)
df_iris['label'] = pd.Series(iris.target)

print(df_iris.head())

sc = SparkContext().getOrCreate()
sqlContext = SQLContext(sc)

data = sqlContext.createDataFrame(df_iris)
print(data.printSchema())

features = iris.feature_names

va = VectorAssembler(inputCols = features, outputCol='features')

va_df = va.transform(data)
va_df = va_df.select(['features', 'label'])
va_df.show(3)

(train, test) = va_df.randomSplit([0.8, 0.2])

rfc = RandomForestClassifier(featuresCol="features", labelCol="label")
rfc = rfc.fit(train)

pred = rfc.transform(test)
pred.show(3)

evaluator=MulticlassClassificationEvaluator(predictionCol="prediction")
acc = evaluator.evaluate(pred)
print("Prediction Accuracy: ", acc)

y_pred=pred.select("prediction").collect()
y_orig=pred.select("label").collect()

cm = confusion_matrix(y_orig, y_pred)
print("Confusion Matrix:")
print(cm)

sc.stop() 

 

References:

1. PySpark random forest classifier