pyspark average(avg) function

In this article, we will show how average function works in PySpark. avg() is an aggregate function which is used to get the average value from the dataframe column/s. We can get average value in three ways.

Lets go through one by one.

First let's create the dataframe for demonstration.

Step 1: Creating a dataframe for the demonstration.

 copy
import pyspark
from pyspark.sql import SparkSession
# create the app name GKINDEX
app = SparkSession.builder.appName('GKINDEX').getOrCreate()

# create grocery data with 5 items with 4 attributes
grocery_data =[{'food_id':112,'item':'onions','cost':234.89,'quantity':4},
               {'food_id':113,'item':'potato','cost':17.39,'quantity':1},
               {'food_id':102,'item':'grains','cost':4234.9,'quantity':84},
                {'food_id':98,'item':'shampoo/soap','cost':10.89,'quantity':2},
                {'food_id':98,'item':'shampoo/soap','cost':100.89,'quantity':20},
               {'food_id':98,'item':'shampoo/soap','cost':1234.89,'quantity':94},
               {'food_id':113,'item':'potato','cost':170.39,'quantity':10},
               {'food_id':113,'item':'potato','cost':34.39,'quantity':2},
               {'food_id':102,'item':'grains','cost':1000.9,'quantity':24},
               {'food_id':56,'item':'oil','cost':134.00,'quantity':10}]

# creating a dataframe from the grocery_data
input_dataframe = app.createDataFrame( grocery_data)

#display
input_dataframe.show()
Output:
 copy
+-------+-------+------------+--------+
|   cost|food_id|        item|quantity|
+-------+-------+------------+--------+
| 234.89|    112|      onions|       4|
|  17.39|    113|      potato|       1|
| 4234.9|    102|      grains|      84|
|  10.89|     98|shampoo/soap|       2|
| 100.89|     98|shampoo/soap|      20|
|1234.89|     98|shampoo/soap|      94|
| 170.39|    113|      potato|      10|
|  34.39|    113|      potato|       2|
| 1000.9|    102|      grains|      24|
|  134.0|     56|         oil|      10|
+-------+-------+------------+--------+

Method - 1 : Using select() method

select() method is used to select the average value from the dataframe columns. It can take single or multipe columns at a time. It will take avg() function as parameter.
But,we have to import avg function from pyspark.sql.functions
Syntax:
 copy
dataframe.select(avg('column1'),............,avg('column n'))
where,
1. dataframe is the input PySpark DataFrame
2. column  specifies the average value to be returned
Example:
In this example will use avg function on cost and quantity columns.
 copy
import pyspark
from pyspark.sql import SparkSession

#import avg function 
from pyspark.sql.functions import avg

# create the app name GKINDEX
app = SparkSession.builder.appName('GKINDEX').getOrCreate()

# create grocery data with 5 items with 4 attributes
grocery_data =[{'food_id':112,'item':'onions','cost':234.89,'quantity':4},
               {'food_id':113,'item':'potato','cost':17.39,'quantity':1},
               {'food_id':102,'item':'grains','cost':4234.9,'quantity':84},
                {'food_id':98,'item':'shampoo/soap','cost':10.89,'quantity':2},
                {'food_id':98,'item':'shampoo/soap','cost':100.89,'quantity':20},
               {'food_id':98,'item':'shampoo/soap','cost':1234.89,'quantity':94},
               {'food_id':113,'item':'potato','cost':170.39,'quantity':10},
               {'food_id':113,'item':'potato','cost':34.39,'quantity':2},
               {'food_id':102,'item':'grains','cost':1000.9,'quantity':24},
               {'food_id':56,'item':'oil','cost':134.00,'quantity':10}]

# creating a dataframe from the grocery_data
input_dataframe = app.createDataFrame( grocery_data)

#get the average of cost and quantity  column
input_dataframe.select(avg('cost'),avg('quantity')).show()
Output:
 copy
+-----------------+-------------+
|        avg(cost)|avg(quantity)|
+-----------------+-------------+
|717.3530000000001|         25.1|
+-----------------+-------------+

Method - 2 : Using agg() method

agg() stands for aggragation which is used to select the average value from the dataframe columns. It will take a dictinary as a parameter in which key will be the column name in the dataframe and value represents the aggregate function name that is avg.we can specify multiple columns to apply the aggregate function

Syntax:
 copy
dataframe.agg({'column1': 'avg',......,'column n':'avg'})where,
1. dataframe is the input PySpark DataFrame
2. column  specifies the average value to be returned
Example:
In this example will use avg function on cost and quantity columns.
 copy
import pyspark
from pyspark.sql import SparkSession

# create the app name GKINDEX
app = SparkSession.builder.appName('GKINDEX').getOrCreate()

# create grocery data with 5 items with 4 attributes
grocery_data =[{'food_id':112,'item':'onions','cost':234.89,'quantity':4},
               {'food_id':113,'item':'potato','cost':17.39,'quantity':1},
               {'food_id':102,'item':'grains','cost':4234.9,'quantity':84},
                {'food_id':98,'item':'shampoo/soap','cost':10.89,'quantity':2},
                {'food_id':98,'item':'shampoo/soap','cost':100.89,'quantity':20},
               {'food_id':98,'item':'shampoo/soap','cost':1234.89,'quantity':94},
               {'food_id':113,'item':'potato','cost':170.39,'quantity':10},
               {'food_id':113,'item':'potato','cost':34.39,'quantity':2},
               {'food_id':102,'item':'grains','cost':1000.9,'quantity':24},
               {'food_id':56,'item':'oil','cost':134.00,'quantity':10}]

# creating a dataframe from the grocery_data
input_dataframe = app.createDataFrame( grocery_data)

#get the average of cost and quantity column
input_dataframe.agg({'cost': 'avg','quantity':'avg'}).show()


Output:
 copy
+-----------------+-------------+
|        avg(cost)|avg(quantity)|
+-----------------+-------------+
|717.3530000000001|         25.1|
+-----------------+-------------+

Method - 3 : Using groupBy() with avg()

If we want to get the average based on values in a group we have to use groupBy() function.
This will group the values which are similar in a column and return the average based on group.
Syntax:
 copy
dataframe.groupBy('group_column').avg('column')
where,
1. dataframe is the input dataframe
2. group_column is the column where values are grouped
3. column is the column name to get average value based on group_column
Example:
Python program to get average by grouping the item column with cost
 copy
import pyspark
from pyspark.sql import SparkSession

# create the app name GKINDEX
app = SparkSession.builder.appName('GKINDEX').getOrCreate()

# create grocery data with 5 items with 4 attributes
grocery_data =[{'food_id':112,'item':'onions','cost':234.89,'quantity':4},
               {'food_id':113,'item':'potato','cost':17.39,'quantity':1},
               {'food_id':102,'item':'grains','cost':4234.9,'quantity':84},
                {'food_id':98,'item':'shampoo/soap','cost':10.89,'quantity':2},
                {'food_id':98,'item':'shampoo/soap','cost':100.89,'quantity':20},
               {'food_id':98,'item':'shampoo/soap','cost':1234.89,'quantity':94},
               {'food_id':113,'item':'potato','cost':170.39,'quantity':10},
               {'food_id':113,'item':'potato','cost':34.39,'quantity':2},
               {'food_id':102,'item':'grains','cost':1000.9,'quantity':24},
               {'food_id':56,'item':'oil','cost':134.00,'quantity':10}]

# creating a dataframe from the grocery_data
input_dataframe = app.createDataFrame( grocery_data)

#get the average of cost column groued by item
input_dataframe.groupBy('item').avg('cost').show()
Output:
 copy
+------------+------------------+
|        item|         avg(cost)|
+------------+------------------+
|      grains|2617.8999999999996|
|      onions|            234.89|
|      potato| 74.05666666666666|
|shampoo/soap|448.89000000000004|
|         oil|             134.0|
+------------+------------------+