for the question 8 I wrote following code:

def check(data):
    answ = np.array([False for i in range(data.shape[0])])
    for x in range(1,data.shape[1]-1):
        answ = answ | np.array(data[:,x]==0) 
    return answ
print(data[check(data)].shape[0])

which gives me answer of 638. this answer was not in the list, so I choose randomly one of them. After that I checked your code and found that last column ("Outcome") is not participate in the calculations. I assumed that "Outcome" is also feature which may be unknown. So my question is why we do not use nan values for unknown values instead of 0-s?

P.S. Sorry for my English