statistics.py

from pyspark.sql.functions import col,avg,min
# Min num implicit ratings for a song
print("Minimum implicit ratings for a song: ")
msd.filter(col("num_plays") > 0).groupBy("songId").count().select(min("count")).show()

# Avg num implicit ratings per songs
print("Average implicit ratings per song: ")
msd.filter(col("num_plays") > 0).groupBy("songId").count().select(avg("count")).show()

# Min num implicit ratings from a user
print("Minimum implicit ratings from a user: ")
msd.filter(col("num_plays") > 0).groupBy("userId").count().select(min("count")).show()

# Avg num implicit ratings for users
print("Average implicit ratings per user: ")
msd.filter(col("num_plays") > 0).groupBy("userId").count().select(avg("count")).show()

In this exercise, we are going to combine the .groupBy() and .filter() methods that you've used previously to calculate the min() and avg() number of users that have rated each song, and the min() and avg() number of songs that each user has rated.Because our data now includes 0's for items not yet consumed, we'll need to .filter() them out when doing grouped summary statistics like this. The msd dataset is provided for you here.

Great work. Users have at least 21 implicit ratings with an average of 77 and each song has at least 3 implicit ratings with an average of 35.