Python Spark take random sample based on column

Question

This is my python spark code

def parseLinesEcf4(line): #get the fields we need
    fields = line.split('\t')
    id1 = fields[0]
    id2 = fields[1]
    ecfp4 = float(fields[2])
    return (id1, id2, ecfp4) #return two fields

conf = SparkConf().setMaster("local").setAppName("Second")
sc = SparkContext(conf = conf)
fileTwo = sc.textFile("PS21_ECFP4.tsv") #loads the data
dataTwo = fileTwo.map(parseLinesEcf4)

My input looks like this

and the size of my file is around 900GB. What I need is to take the rows of which unique values of column 1 correspond to 10% of the unique values of the same column, because one compound has more than one entries.

I tried takeSampe() and sampleBy() but both don't return what I am looking for.

Any help??

Answer 1

You can try to use pyspark.ml library.

from pyspark.ml.evaluation import RegressionEvaluator
from pyspark.ml.regression import LinearRegression
from pyspark.ml.tuning import ParamGridBuilder, TrainValidationSplit

# Prepare training and test data.
data = spark.read.format("libsvm")\
    .load("data/mllib/sample_linear_regression_data.txt")
train, test = data.randomSplit([0.9, 0.1], seed=12345)

https://spark.apache.org/docs/2.1.0/ml-tuning.html#example-model-selection-via-train-validation-split

But be aware, to use it you need to vectorise your data using VectorAssembler

from pyspark.ml.linalg import Vectors
from pyspark.ml.feature import VectorAssembler

dataset = spark.createDataFrame(
    [(0, 18, 1.0, Vectors.dense([0.0, 10.0, 0.5]), 1.0)],
    ["id", "hour", "mobile", "userFeatures", "clicked"])

assembler = VectorAssembler(
    inputCols=["hour", "mobile", "userFeatures"],
    outputCol="features")

output = assembler.transform(dataset)
print("Assembled columns 'hour', 'mobile', 'userFeatures' to vector column 'features'")
output.select("features", "clicked").show(truncate=False)

https://spark.apache.org/docs/latest/ml-features.html#vectorassembler