don't know how u r gonna generate train/test, but stack() function in pandas should be useful

Good day,

You can try this code snippet. I did it manually through looping. Hope it answers your question

SubFile = pd.read_csv('SampleSubmission.csv') #Reading submissionsample file

probs = model.predict(x_test) #predicting x_test in the normal shape 10000 rows

IDXPCODE = list(SubFile.iloc[:,0]) #Extracting product names from SubFile

Labels = []

for i in range(test.shape[0]):

for j in range(21):

Labels.append(probs[i][j])

len(Labels) #make sure len is 210000 records

Submission = pd.DataFrame(list(zip(IDXPCODE,Label)),columns = ['ID X PCODE', 'Label'])

Submission.head(42)

thanks @kejiaq and @tichavona

It also depends on the approach you take to solving the problem. You can melt the train and test features then generate the IDXPCODE column.