f4

###f4:

pr=graphlab.SFrame('1gb.csv')


## filter based on gender
pr=pr[pr['Gender']!='0']

len(pr)

Out[11]:

1048575

In [12]:

len(pr1)

Out[12]:

472767

pr1=pr1[pr1['Age']!=0]


pr1['sentiment']= pr1['Outcome']=='w'


num=len(pr1.filter_by([1], 'sentiment'))

pr1=pr1.topk('sentiment', k=3*num, reverse=False)


pr1.save('data/1gb_filtered.csv', format='csv')

train_data,test_data = pr1.random_split(.8)



idn=pr1[pr1['Country']=='IDN']

In [39]:

usa=pr1[pr1['Country']=='USA']

In [40]:

len(idn)

Out[40]:

24463

In [41]:

len(usa)

Out[41]:

19143

train_data_usa,test_data_usa = usa.random_split(.8)


usa=pr1[pr1['Country']=='USA']
train_data_android,test_data_android = android.random_split(.8)


model=graphlab.logistic_classifier.create(train_data_android,target='sentiment',features=['CreativeType','ExchangeBid'],validation_set=test_data_android)