Shuffle rows of a large csv-CodePudding

I want to shuffle this dataset to have a random set. It has 1.6 million rows but the first are 0 and the last 4, so I need pick samples randomly to have more than one class. The actual code prints only class 0 (meaning in just 1 class). I took advice from this platform but doesn't work.

fid = open("sentiment_train.csv", "r")

li = fid.readlines(16000000)


random.shuffle(li)

fid2 = open("shuffled_train.csv", "w")

fid2.writelines(li)

fid2.close()

fid.close()

sentiment_onefourty_train = pd.read_csv('shuffled_train.csv', header= 0, delimiter=",", usecols=[0,5], nrows=100000)

sentiment_onefourty_train.columns=['target', 'text']

print(sentiment_onefourty_train['target'].value_counts())

CodePudding user response：

Because you read in your data using Pandas, you can also do the randomisation in a different way using pd.sample:

df = pd.read_csv('sentiment_train.csv', header= 0, delimiter=",", usecols=[0,5])
df.columns=['target', 'text']
df1 = df.sample(n=100000)

If this fails, it might be good to check out the amount of unique values and how frequent they appear. If the first 1,599,999 are 0 and the last is only 4, then the chances are that you won't get any 4.