Question

我有以下通用格式的数据，我想重新采样到30天的时间序列窗口：</ p>

'customer_id','transaction_dt','product','price','units'
1,2004-01-02,thing1,25,47
1,2004-01-17,thing2,150,8
2,2004-01-29,thing2,150,25
3,2017-07-15,thing3,55,17
3,2016-05-12,thing3,55,47
4,2012-02-23,thing2,150,22
4,2009-10-10,thing1,25,12
4,2014-04-04,thing2,150,2
5,2008-07-09,thing2,150,43

我希望30天的窗口能够在2014-01-01开始，并以12-31-2018结束。不保证每个客户都会在每个窗口中都有记录。如果客户在一个窗口中有多个交易，那么它需要加权平均价格，对单位求和，并连接产品名称，以便为每个窗口的每个客户创建一条记录。

到目前为止我所拥有的是这样的：

wa = lambda x:np.average(x, weights=df.loc[x.index, 'units'])
con = lambda x: '/'.join(x))

agg_funcs = {'customer_id':'first',
             'product':'con',
             'price':'wa',
             'transaction_dt':'first',
             'units':'sum'}

df_window = df.groupby(['customer_id', pd.Grouper(freq='30D')]).agg(agg_funcs)
df_window_final = df_window.unstack('customer_id', fill_value=0)

如果有人知道一些更好的方法来解决这个问题（特别是使用就地和/或矢量化方法），我将不胜感激。理想情况下，我还想将窗口的开始和停止日期添加为行的列。

最终输出看起来像这样：

'customer_id','transaction_dt','product','price','units','window_start_dt','window_end_dt'
1,2004-01-02,thing1/thing2,(weighted average price),(total units),(window_start_dt),(window_end_dt)
2,2004-01-29,thing2,(weighted average price),(total units),(window_start_dt),(window_end_dt)
3,2017-07-15,thing3,(weighted average price),(total units),(window_start_dt),(window_end_dt)
3,2016-05-12,thing3,(weighted average price),(total units),(window_start_dt),(window_end_dt)
4,2012-02-23,thing2,(weighted average price),(total units),(window_start_dt),(window_end_dt)
4,2009-10-10,thing1,(weighted average price),(total units),(window_start_dt),(window_end_dt)
4,2014-04-04,thing2,(weighted average price),(total units),(window_start_dt),(window_end_dt)
5,2008-07-09,thing2,(weighted average price),(total units),(window_start_dt),(window_end_dt)

Answer 1

编辑新解决方案。我认为您可以将每个transaction_dt转换为30天的Period对象，然后进行分组。

p = pd.period_range('2004-1-1', '12-31-2018',freq='30D')
def find_period(v):
    p_idx = np.argmax(v < p.end_time)
    return p[p_idx]
df['period'] = df['transaction_dt'].apply(find_period)
df

   customer_id transaction_dt product  price  units     period
0            1     2004-01-02  thing1     25     47 2004-01-01
1            1     2004-01-17  thing2    150      8 2004-01-01
2            2     2004-01-29  thing2    150     25 2004-01-01
3            3     2017-07-15  thing3     55     17 2017-06-21
4            3     2016-05-12  thing3     55     47 2016-04-27
5            4     2012-02-23  thing2    150     22 2012-02-18
6            4     2009-10-10  thing1     25     12 2009-10-01
7            4     2014-04-04  thing2    150      2 2014-03-09
8            5     2008-07-09  thing2    150     43 2008-07-08

我们现在可以使用此数据框来获取产品的连接，加权平均价格和单位总和。然后我们使用一些Period功能来获得结束时间。

def my_funcs(df):
    data = {}
    data['product'] = '/'.join(df['product'].tolist())
    data['units'] = df.units.sum()
    data['price'] = np.average(df['price'], weights=df['units'])
    data['transaction_dt'] = df['transaction_dt'].iloc[0]
    data['window_start_time'] = df['period'].iloc[0].start_time
    data['window_end_time'] = df['period'].iloc[0].end_time
    return pd.Series(data, index=['transaction_dt', 'product', 'price','units', 
                                  'window_start_time', 'window_end_time'])

df.groupby(['customer_id', 'period']).apply(my_funcs).reset_index('period', drop=True)

Python PANDAS：使用Groupby重新采样多变量时间序列

1 个答案: