Python中的数据分析（数据框和嵌套循环）答案

【问题标题】：Analysis of Data in Python (Dataframe and nested loops)Python中的数据分析（数据框和嵌套循环）
【发布时间】：2018-09-03 16:19:32
【问题描述】：

您好，我是新手，想了解如何在 python 中使用嵌套循环。我试图理解总结相同的值并学会使用 group_by 函数（基于我今天看到的 stackoverflow 中的另一个问题）。我想学习 pytonic-dataframe 方式。

现在我想用以下方式总结工作日。我根据场景来总结单位，例如：场景 = 1，公司 = A，国家 = 美国，单位 = HR+企业客户，总结工作时间 = 65+63 = 128 等。在原始数据之后我包括输出应该是什么样子。我不确定这是否也适用于 group_by 这更像是一种枢轴方式。

我从嵌套循环开始，但在索引日期时遇到了问题。因此，我的代码仅按日期过滤，效率不高，但有效。我了解到嵌套循环对于数据帧来说是不够的，但不确定我可以走哪条路。代码如下所示：

import pandas as pd

working_date_start = '2017-07-14'
working_date_end = '2017-07-15'
flag_scenario = 0 
Scenario = 0

df = pd.read_csv('C:/Comapny_WorkingHours.csv', encoding='cp1252', sep=';', index_col=None).dropna()

df = df[(df['working_date'] >= working_date_start) & (df['working_date'] < working_date_end) & (df['flag'] == flag_scenario) & (df['Scenario'] >= Scenario)]

pd_date = pd.DatetimeIndex(df['working_date'].values)
df['working_date'] = pd_date
index_data = df.set_index('working_date')

for current_date in index_data.index.unique():
     print('calculating date: ' +str(current_date))

     for i in range(0, len(df)):

         for j in range(i+1, len(df)):

             if df.iloc[i]['Scenario'] == df.iloc[j]['Scenario'] and df.iloc[i]['Unit'] != df.iloc[j]['Unit'] and df.iloc[i]['Company'] == 'Company A' and df.iloc[j]['Company'] == 'Company A' and df.iloc[i]['Country'] == 'USA' and df.iloc[j]['Country'] == 'USA':
                 print(df.iloc[i]['Scenario'], df.iloc[j]['Scenario'])
                 print(df.iloc[i]['Unit'], df.iloc[j]['Unit'])

原始数据如下所示：

working_date    flag    Scenario    Working Hours   Country Company Unit
2017-07-14  0   1   65  USA Company A   HR
2017-07-14  0   2   75  USA Company A   HR
2017-07-14  0   3   73  USA Company A   HR
2017-07-14  0   4   66  USA Company A   HR
2017-07-14  0   1   63  USA Company A   Corporate Client
2017-07-14  0   2   51  USA Company A   Corporate Client
2017-07-14  0   3   60  USA Company A   Corporate Client
2017-07-14  0   4   55  USA Company A   Corporate Client
2017-07-14  0   1   71  USA Company A   Controlling
2017-07-14  0   2   45  USA Company A   Controlling
2017-07-14  0   3   76  USA Company A   Controlling
2017-07-14  0   4   62  USA Company A   Controlling
2017-07-14  0   1   57  USA Company A   Corporate Center
2017-07-14  0   2   64  USA Company A   Corporate Center
2017-07-14  0   3   68  USA Company A   Corporate Center
2017-07-14  0   4   69  USA Company A   Corporate Center
2017-07-14  0   1   54  USA Company B   Private and Business Customers
2017-07-14  0   2   62  USA Company B   private and business customers
2017-07-14  0   3   47  USA Company B   private and business customers
2017-07-14  0   4   62  USA Company B   private and business customers
2017-07-14  0   1   45  USA Company B   Marketing
2017-07-14  0   2   78  USA Company B   Marketing
2017-07-14  0   3   59  USA Company B   Marketing
2017-07-14  0   4   78  USA Company B   Marketing
2017-07-14  0   1   49  USA Company B   IT
2017-07-14  0   2   74  USA Company B   IT
2017-07-14  0   3   78  USA Company B   IT
2017-07-14  0   4   55  USA Company B   IT
2017-07-14  0   1   66  USA Company B   Project Management
2017-07-14  0   2   76  USA Company B   Project Management
2017-07-14  0   3   53  USA Company B   Project Management
2017-07-14  0   4   58  USA Company B   Project Management
2017-07-15  0   1   56  USA Company A   HR
2017-07-15  0   2   54  USA Company A   HR
2017-07-15  0   3   77  USA Company A   HR
2017-07-15  0   4   58  USA Company A   HR
2017-07-15  0   1   78  USA Company A   Corporate Client
2017-07-15  0   2   76  USA Company A   Corporate Client
2017-07-15  0   3   59  USA Company A   Corporate Client
2017-07-15  0   4   56  USA Company A   Corporate Client
2017-07-15  0   1   57  USA Company A   Controlling
2017-07-15  0   2   54  USA Company A   Controlling
2017-07-15  0   3   56  USA Company A   Controlling
2017-07-15  0   4   74  USA Company A   Controlling
2017-07-15  0   1   71  USA Company A   Corporate Center
2017-07-15  0   2   75  USA Company A   Corporate Center
2017-07-15  0   3   79  USA Company A   Corporate Center
2017-07-15  0   4   78  USA Company A   Corporate Center
2017-07-15  0   1   74  USA Company B   Private and Business Customers
2017-07-15  0   2   72  USA Company B   private and business customers
2017-07-15  0   3   66  USA Company B   private and business customers
2017-07-15  0   4   66  USA Company B   private and business customers
2017-07-15  0   1   69  USA Company B   Marketing
2017-07-15  0   2   69  USA Company B   Marketing
2017-07-15  0   3   63  USA Company B   Marketing
2017-07-15  0   4   59  USA Company B   Marketing
2017-07-15  0   1   57  USA Company B   IT
2017-07-15  0   2   67  USA Company B   IT
2017-07-15  0   3   77  USA Company B   IT
2017-07-15  0   4   60  USA Company B   IT
2017-07-15  0   1   55  USA Company B   Project Management
2017-07-15  0   2   57  USA Company B   Project Management
2017-07-15  0   3   80  USA Company B   Project Management
2017-07-15  0   4   59  USA Company B   Project Management

我想要的输出如下所示：

working_date    Scenario   Units                                Working Hours Summed Up
2017-07-14      1          HR_Corporate Client                  128
2017-07-14      1          HR_Controlling                       136
2017-07-14      1          HR_Corporate Center                  122
2017-07-14      2          HR_Corporate Client                  126
2017-07-14      2          HR_Controlling                       120
2017-07-14      2          HR_Corporate Center                  139
2017-07-14      3          HR_Corporate Client                  133
2017-07-14      3          HR_Controlling                       149
2017-07-14      3          HR_Corporate Center                  141
2017-07-14      4          HR_Corporate Client                  121
2017-07-14      4          HR_Controlling                       128
2017-07-14      4          HR_Corporate Center                  135
2017-07-14      1          Corporate Client_Controlling         134
2017-07-14      1          Corporate Client_Corporate Center    120
2017-07-14      2          Corporate Client_Controlling          96
2017-07-14      2          Corporate Client_Corporate Center    115
2017-07-14      3          Corporate Client_Controlling         136
2017-07-14      3          Corporate Client_Corporate Center    128
2017-07-14      4          Corporate Client_Controlling         117
2017-07-14      4          Corporate Client_Corporate Center    124
2017-07-14      1          Controlling_Corporate Center         128
2017-07-14      2          Controlling_Corporate Center         109
2017-07-14      3          Controlling_Corporate Center         144
2017-07-14      4          Controlling_Corporate Center         131

【问题讨论】：

标签： python pandas dataframe nested-loops

【解决方案1】：

import pandas as pd
df = pd.read_csv('C:/Comapny_WorkingHours.csv', encoding='cp1252', sep=';', index_col=None).dropna()

df = df.reset_index(drop=False)

# this will give you the unique combinations of two units
from itertools import combinations 
scenario_list = df['Scenario'].unique().tolist()

# this creates a dict containing the scene and corresponidng units combos
combos_dict = {}
for scene in scenario_list:
    units_list = df[df['Scenario'] == scene]['Unit'].unique().tolist()
    combos_dict[scene] = list(combinations(units_list, 2))

new_df = pd.DataFrame() 
for key in combos_dict.keys():
    # filters the dataframe by the scenario matched in the combo_dict
    filter_df = df[df['Scenario'] == key]
    for combo in combos_dict[key]:
        # itterates through the combo_dict values to create a sub_filter
        # that is used to create a new final dataframe 
        sub_filter = filter_df[(filter_df['Unit'] == combo[0]) | 
                               (filter_df['Unit'] == combo[1])] 
        sub_df = pd.DataFrame(data=[[sub_filter['working_date'].iloc[0],
                                     key,
                                     '{}_{}'.format(combo[0], combo[1]),
                                     sum(sub_filter['Working Hours'])]],
                              columns=['working_date',
                                       'Scenario',
                                       'Units',                                
                                       'Working Hours Summed Up'])
        # creates a new dataframe with the desired output
        new_df = new_df.append(sub_df)

【讨论】：

非常感谢，但不知何故，我想收到的结果有所不同。我需要使用不同单位的组合来总结工作时间。看看我作为例子给出的输出。希望它更清楚？
我已经更新了代码，希望这能更准确地反映您的预期结果
非常感谢。我收到一条错误消息 = AttributeError: 'DataFrame' object has no attribute 'unique' for that line 'units_list = df[df['Scenario'] == scene].unique().tolist()'
你要重新格式化为units_list = df[df['Scenario'] == scene]['Unit'].unique().tolist()，这是因为当我创建了unit_list，尽管过滤了数据框，但我从未真正指定过该列...松散的错误:)
你是我的新英雄 :)。非常感谢。你能帮我一个忙，并评论一下我理解你所做的代码吗？再次感谢！！！