如何使用每个元组的第一个值作为键将六个元组列表连接到一个数据帧中？

Question

我正在测试一个具有api的服务，该api可以从中提取解析的10K公司数据。对于提取的每个指标（息税前利润、现金、总资产等），我将季度日期和指标存储在一个元组中，并将每个元组存储在一个列表中。结果是6个43-80元组的列表。我想要一个带有公司股票代码、日期和指标列的数据框架。我如何将我拥有的（元组列表）转换成那个

下面的代码用于提取数据（这是示例，因此不收费）：

import numpy as np
import json
import pandas as pd

content = requests.get(r'https://eodhistoricaldata.com/api/fundamentals/AAPL.US?api_token=OeAFFmMliFG5orCUuwAKQ8l4WWFQ67YX')

ebit_list = []
date_list = []
totalassets_list = []
cash_list = []
totalCurrentAssets_list = []
totalCurrentLiabilities_list = []


for i in content.json()['Financials']['Income_Statement']['quarterly']:

    try:
        ebit_list.append((i, float(content.json()['Financials']['Income_Statement']['quarterly'][i]['ebit'])))
    except:
        pass

    try:
        date_list.append(i)
    except:
        pass

    try:
        totalassets_list.append((i, float(content.json()['Financials']['Balance_Sheet']['quarterly'][i]['totalAssets'])))
    except:
        pass



for i in content.json()['Financials']['Balance_Sheet']['quarterly']:
    #print(i, float(content.json()['Financials']['Balance_Sheet']['quarterly']['2019-12-28']['totalCurrentLiabilities']))
    try:
        cash_list.append((i, float(content.json()['Financials']['Balance_Sheet']['quarterly'][i]['cash'])))
    except:
        pass

    try:
        totalCurrentAssets_list.append((i, float(content.json()['Financials']['Balance_Sheet']['quarterly'][i]['totalCurrentAssets'])))
    except:
        pass

    try:
        totalCurrentLiabilities_list.append((i, float(content.json()['Financials']['Balance_Sheet']['quarterly'][i]['totalCurrentLiabilities'])))
    except:
        pass

我想要一个包含所有日期的数据框（意味着如果缺少一个度量，则填写一个零）和以下列：

date，ebit，totalassets，cash，totalCurrentAssets，totalCurrentLiabilities

我不知道如何提取元组和每个元组中的值

Answer 1

您可以使用pandas.Series中的^{}函数将日期与所需的数据进行匹配。这将为没有匹配值的单元格插入NaN，这将使以后更容易处理丢失的数据。如果仍然要填充零，可以使用^{}

# Create a dataframe using date
df = pd.DataFrame({'date': date_list})

# To avoid the code getting messy in the next steps
stuff = {'ebit': ebit_list, 'totalassets': totalassets_list, 'cash': cash_list, 'totalCurrentAssets': totalCurrentAssets_list, 'totalCurrentLiabilities': totalCurrentLiabilities_list}

for name, values in stuff.items():
    value_dict = {t[0]: t[1] for t in values}   # t is each tuple in the list
    df[name] = df['date'].map(value_dict)       # map will match the correct date to the value 

# assuming you need the dataframe to be sorted by date
df['date'] = pd.to_datetime(df['date'])         # we should use actual numbers instead of date string
df.sort_values('date', inplace=True, ignore_index=True)

# if you want to fill 0s to missing values
# df.fillna(0, inplace=True)

ignore_index中的sort_values参数用于确保排序后索引不会混乱。如果pandas版本是旧版本，则排序时可能会给出TypeError: sort_values() got an unexpected keyword argument 'ignore_index'。如果是这样的话，您应该使用以下方法来重置索引

df.sort_values('date', inplace=True)
df.reset_index(inplace=True)

最后是df

         date          ebit   totalassets          cash  totalCurrentAssets  totalCurrentLiabilities
0  2000-03-31           NaN  7.007000e+09           NaN                 NaN             1.853000e+09
1  2000-06-30           NaN  6.932000e+09           NaN                 NaN             1.873000e+09
2  2000-09-30           NaN  6.803000e+09           NaN                 NaN             1.933000e+09
3  2000-12-31  0.000000e+00  5.986000e+09           NaN                 NaN             1.637000e+09
4  2001-03-31  0.000000e+00  6.130000e+09           NaN                 NaN             1.795000e+09
..        ...           ...           ...           ...                 ...                      ...
75 2018-12-29  2.334600e+10  3.737190e+11  4.477100e+10        1.408280e+11             1.082830e+11
76 2019-03-30  1.341500e+10  3.419980e+11  3.798800e+10        1.233460e+11             9.377200e+10
77 2019-06-29  1.154400e+10  3.222390e+11  5.053000e+10        1.349730e+11             8.970400e+10
78 2019-09-28  1.562500e+10  3.385160e+11  4.884400e+10        1.628190e+11             1.057180e+11
79 2019-12-28  2.556900e+10  3.406180e+11  3.977100e+10        1.632310e+11             1.021610e+11

Answer 2

我无法让您的示例工作，请求未定义

但这里有一些代码可以满足您的要求：

#!/usr/bin/env python
# -*- coding: utf-8 -*-

import pandas as pd


def create_df(list_of_lists):
    pd.DataFrame({x[0]: pd.Series(x[1:]) for x in list of lists})

Answer 3

实际上，我们可以大大简化此代码，以获得所需的结果（并使其在将来更易于调整！）

完成的代码在这里，更详细的解释如下：

import numpy as np
import json
import pandas as pd
import requests

content = requests.get(r'https://eodhistoricaldata.com/api/fundamentals/AAPL.US?api_token=OeAFFmMliFG5orCUuwAKQ8l4WWFQ67YX')

income_data = content.json()['Financials']['Income_Statement']['quarterly']
income = pd.DataFrame.from_dict(income_data).transpose().set_index("date")
income = income[['ebit']]

balance_data = content.json()['Financials']['Balance_Sheet']['quarterly']
balance = pd.DataFrame.from_dict(balance_data).transpose().set_index("date")
balance = balance[['totalAssets', 'cash', 'totalCurrentAssets', 'totalCurrentLiabilities']]

financials = income.merge(balance, left_index = True, right_index = True).fillna(0)

财务数据框架如下所示（仅显示2005-2009年的数据）：

| date       |      ebit |   totalAssets |       cash |   totalCurrentAssets |   totalCurrentLiabilities |
|:     -|     :|       :|     -:|          -:|             :|
| 2009-12-26 | 4.758e+09 |    5.3926e+10 | 7.609e+09  |           3.3332e+10 |                1.3097e+10 |
| 2009-09-26 | 0         |    4.7501e+10 | 5.263e+09  |           3.1555e+10 |                1.1506e+10 |
| 2009-06-27 | 1.732e+09 |    4.814e+10  | 5.605e+09  |           3.517e+10  |                1.6661e+10 |
| 2009-03-31 | 0         |    4.3237e+10 | 4.466e+09  |           0          |                1.3751e+10 |
| 2008-12-31 | 0         |    4.2787e+10 | 7.236e+09  |           0          |                1.4757e+10 |
| 2008-09-30 | 0         |    3.9572e+10 | 1.1875e+10 |           0          |                1.4092e+10 |
| 2008-06-30 | 0         |    3.1709e+10 | 9.373e+09  |           0          |                9.218e+09  |
| 2008-03-31 | 0         |    3.0471e+10 | 9.07e+09   |           0          |                9.634e+09  |
| 2007-12-31 | 0         |    3.0039e+10 | 9.162e+09  |           0          |                1.0535e+10 |
| 2007-09-30 | 0         |    2.5347e+10 | 9.352e+09  |           0          |                9.299e+09  |
| 2007-06-30 | 0         |    2.1647e+10 | 7.118e+09  |           0          |                6.992e+09  |
| 2007-03-31 | 0         |    1.8711e+10 | 7.095e+09  |           0          |                5.485e+09  |
| 2006-12-31 | 0         |    1.9461e+10 | 7.159e+09  |           0          |                7.337e+09  |
| 2006-09-30 | 0         |    1.7205e+10 | 6.392e+09  |           0          |                6.471e+09  |
| 2006-06-30 | 0         |    1.5114e+10 | 0          |           0          |                5.023e+09  |
| 2006-03-31 | 0         |    1.3911e+10 | 0          |           0          |                4.456e+09  |
| 2005-12-31 | 0         |    1.4181e+10 | 0          |           0          |                5.06e+09   |
| 2005-09-30 | 0         |    1.1551e+10 | 3.491e+09  |           0          |                3.484e+09  |
| 2005-06-30 | 0         |    1.0488e+10 | 0          |           0          |                3.123e+09  |
| 2005-03-31 | 0         |    1.0111e+10 | 0          |           0          |                3.352e+09  |

---

content.json()['Financials']['Income_Statement']['quarterly']的结果是一个字典，每个键都是日期，每个值都是第二个字典，其中包含列数据

{'2005-03-31': {'date': '2005-03-31',
                'filing_date': None,
                'currency_symbol': 'USD',
                'researchDevelopment': '120000000.00',
                ...},
'2005-06-30': {...},
...}

由于是这种情况，您实际上可以使用

pd.DataFrame.from_dict(income_data).transpose().set_index("date")

由于JSON的结构，转置是必要的。Pandas需要一个格式为{'column name': data}的字典。由于键是日期，您将首先获得一个数据框，其中行标记为“totalAssets”、“cash”等，列为日期。transpose()命令翻转行和列，使其符合您需要的格式最后一个.set_index("date")命令用于使用“日期”数据而不是初始键日期，以保持一致性并命名索引。它是完全可选的

现在，这个数据框架将包含JSON文件中的每一列，但您只对其中的几列感兴趣。代码

income = income[['ebit']]

仅从数据中选择相关列

由于要从两个不同的源提取数据，因此确实需要创建两个不同的表。这还有一个额外的好处，那就是你可以更清楚地看到哪些栏目是从“损益表”中提取出来的，哪些栏目是从“资产负债表”中提取出来的

最后一行

financials = income.merge(balance, left_index = True, right_index = True).fillna(0)

使用索引（在本例中为“日期”列）将两个表合并在一起fillna(0)确保按照您的请求，用零值替换任何缺失的数据

如果您最终需要添加另一个表，例如“现金流”，您可以使用相同的代码行创建该表并选择相关列，然后添加第二个合并行：

cashflow_data = content.json()['Financials']['Balance_Sheet']['quarterly']
cashflow = pd.DataFrame.from_dict(cashflow_data).transpose().set_index("date")
cashflow = cashflow[['accountsPayable', 'liabilitiesAndStockholdersEquity']]
...
financials.merge(cashflow, left_index = True, right_index = True).fillna(0)

---

作为一个额外提示，源JSON中有相当多的数据！要查看任何给定表中的可用列，请使用以下命令：

cashflow.columns.sort_values()

要获取按字母顺序排列的列列表，可以使用：

      ['accountsPayable', 'accumulatedAmortization', 'accumulatedDepreciation',
       'accumulatedOtherComprehensiveIncome', 'additionalPaidInCapital',
       'capitalLeaseObligations', 'capitalSurpluse', 'cash',
       'cashAndShortTermInvestments', 'commonStock',
       'commonStockSharesOutstanding', 'commonStockTotalEquity',
       'currency_symbol', 'deferredLongTermAssetCharges',
       'deferredLongTermLiab', 'filing_date', 'goodWill', 'intangibleAssets',
       'inventory', 'liabilitiesAndStockholdersEquity', 'longTermDebt',
       'longTermDebtTotal', 'longTermInvestments', 'negativeGoodwill',
       'netReceivables', 'netTangibleAssets', 'nonCurrentAssetsTotal',
       'nonCurrentLiabilitiesOther', 'nonCurrentLiabilitiesTotal',
       'nonCurrrentAssetsOther', 'noncontrollingInterestInConsolidatedEntity',
       'otherAssets', 'otherCurrentAssets', 'otherCurrentLiab', 'otherLiab',
       'otherStockholderEquity', 'preferredStockRedeemable',
       'preferredStockTotalEquity', 'propertyPlantAndEquipmentGross',
       'propertyPlantEquipment', 'retainedEarnings',
       'retainedEarningsTotalEquity', 'shortLongTermDebt', 'shortTermDebt',
       'shortTermInvestments',
       'temporaryEquityRedeemableNoncontrollingInterests', 'totalAssets',
       'totalCurrentAssets', 'totalCurrentLiabilities', 'totalLiab',
       'totalPermanentEquity', 'totalStockholderEquity', 'treasuryStock',
       'warrants']

当数据中出现拼写错误时，如上面的“capitalSurpluse”中，这也非常有用