Python：基於整數值範圍在Pandas數據框中創建組列。

Question

對於diff列中的每個範圍[0, 150]，我希望創建一個組列，每次範圍重置時都會增加1。當diff爲負值時，範圍重置。

import pandas as pd 
df = pd.DataFrame({'year': [2016, 2016, 2016, 2016, 2016, 2016, 2016], 
        'month' : [1, 1, 2, 3, 3, 3, 3], 
        'day': [23, 25, 1, 1, 7, 20, 30]}) 
df = pd.to_datetime(df) 
df = pd.concat([df, pd.Series(data=[15, 35, 80, 5, 20, 45, 90])], axis=1) 
df.columns = ['date', 'percentworn'] 
col_shift = ['percentworn'] 
df_shift = df.shift(1).loc[:, col_shift] 
df_combined = df.join(df_shift, how='inner', rsuffix='_2') 
df_combined.fillna(value=0,inplace=True) 
df_combined['diff'] = df_combined['percentworn'] - df_combined['percentworn_2'] 

的grp列應具有0, 0, 0, 1, 1, 1, 1。我試過的代碼是

def grping(df): 
    df_ = df.copy(deep=True) 
    i = 0 
    if df_['diff'] >= 0: 
     df_['grp'] = i 
    else: 
     i += 1 
     df_['grp'] = i 
    return df_ 
df_combined.apply(grping,axis=1) 

我需要i += 1增加後仍然存在。我怎樣才能做到這一點？還是有更好的方法來獲得理想的結果？

Answer 1

IIUC可以測試'diff'列是否是產生一個布爾陣列否定的，則投這int並且呼叫cumsum：

In [313]: 
df_combined['group'] = (df_combined['diff'] < 0).astype(int).cumsum() 
df_combined 

Out[313]: 
     date percentworn percentworn_2 diff group 
0 2016-01-23   15   0.0 15.0  0 
1 2016-01-25   35   15.0 20.0  0 
2 2016-02-01   80   35.0 45.0  0 
3 2016-03-01   5   80.0 -75.0  1 
4 2016-03-07   20   5.0 15.0  1 
5 2016-03-20   45   20.0 25.0  1 
6 2016-03-30   90   45.0 45.0  1 

打破上述向下：

In [314]: 
df_combined['diff'] < 0 

Out[314]: 
0 False 
1 False 
2 False 
3  True 
4 False 
5 False 
6 False 
Name: diff, dtype: bool 

In [316]: 
(df_combined['diff'] < 0).astype(int) 

Out[316]: 
0 0 
1 0 
2 0 
3 1 
4 0 
5 0 
6 0 
Name: diff, dtype: int32