パンダ失敗修正

Question

df=pd.read_csv('./data/reward_original_91011.csv') 

ind0=df[df['DatePay'].isnull()].index 

df.ix[ind0]['ValuePay']=0 

ind1=df[~df['DatePay'].isnull()].index 

df['DatePay'] = pd.to_datetime(df['DatePay']) 

df['DateDue'] = pd.to_datetime(df['DateDue']) 

ind2=df.ix[ind1][((df.ix[ind1]['DatePay']-df.ix[ind1]['DateDue'])/np.timedelta64(1,'h')).astype(int)>24].index 

**df.ix[ind2]['ValuePay']=0** 

誰かが私にいくつかのヒントを与えることができますか？なぜ私は上記の太字コードを正常に変更できますか？

Answer 1

は、私はあなたが必要だと思う：

df = pd.read_csv('./data/reward_original_91011.csv', parse_dates=['DatePay','DateDue']) 

#get boolean mask where NaN 
mask = df['DatePay'].isnull() 

#modify column ValuePay by mask 
df.loc[mask, 'ValuePay'] = 0 

#get another mask 
mask1 = ((df['DatePay'] - df['DateDue'])/np.timedelta64(1,'h')) > 24 

#inverted mask AND mask1 -> modify ValuePay 
df.loc[~mask & mask1, 'ValuePay'] = 0 

サンプル：

d = {'DateDue': [pd.Timestamp('2013-10-06 00:00:00'), pd.Timestamp('2013-10-08 00:00:00'), 
       pd.Timestamp('2013-10-08 00:00:00'), pd.Timestamp('2013-10-08 00:00:00')], 
    'DatePay': [np.NaN, pd.Timestamp('2013-10-06 00:00:00'), 
       pd.Timestamp('2013-10-12 00:00:00'), pd.Timestamp('2013-10-10 00:00:00')]} 
df = pd.DataFrame(d) 

print (df) 
    DateDue DatePay 
0 2013-10-06  NaT 
1 2013-10-08 2013-10-06 
2 2013-10-08 2013-10-12 
3 2013-10-08 2013-10-10 

#get boolean mask where NaN 
mask = df['DatePay'].isnull() 

#modify column ValuePay by mask 
df.loc[mask, 'ValuePay'] = 0 

#get another mask 
mask1 = ((df['DatePay'] - df['DateDue'])/np.timedelta64(1,'h')) > 24 

#inverted mask AND mask1 -> modify ValuePay 
df.loc[~mask & mask1, 'ValuePay'] = 0 
print (df) 
    DateDue DatePay ValuePay 
0 2013-10-06  NaT  0.0 
1 2013-10-08 2013-10-06  NaN 
2 2013-10-08 2013-10-12  0.0 
3 2013-10-08 2013-10-10  0.0