在python中转换类别百分比的数据框

Question

我想转换一个数据框，稍后在 python 中计算图表的百分比。

当前帧是这样的

Post ID	Title	Url	Author	Score	Submission_Date	Total_Num_of_Comments	Permalink	Flair	Selftext	TitleAndText	Word Count
k4nllk	Update: Whassup bro?	https://www.reddit.com/r/GME/comments/k4nllk/update_whassup_bro/	matt_xndever	1	2021-01-01 16:58:48	13	/r/GME/comments/k4nllk/update_whassup_bro/	Hedge Fund Tears	asdasdasd	asdasdasdasd	59.0

flairs 是我想要寻找的类别（超过 40 个）。在一个提交日（我只想查看几天），可以有多个具有不同风格的帖子。这些天赋加起来应该是 100%。

所以我想创建一个这样的数据框：

Submission_Date	Discussion	Due Diligence	Hedge Fund Tears	News
01.01.2021	NaN	NaN	1.0	NaN
03.01.2021	NaN	0.333333	0.666667	NaN

我的图表应该是这样的： Plot stacked (100%) bar chart for multiple categories on multiple dates in Python

有人可以帮我准备吗？

感谢和问候

Answer 1

您可以按如下方式解决问题：

• 迭代唯一日期并为每个日期切片数据框
• 使用 pandas value_counts() 计算每个天赋类别的计数
• 通过除以每个切片的大小来获得份额
• 转置包含要附加的共享的 pandas 系列
• 附加每个日期的份额

这是一个示例输入：

df = {
    'Submission_Date': ['2021-01-01', '2021-03-01', '2021-03-01', '2021-03-01', '2021-04-01', '2021-04-01'],
    'Flair': ['Hedge Fund Tears', 'Hedge Fund Tears', 'Hedge Fund Tears', 'Due Diligence', 'Discussion', 'News']
}
df = pd.DataFrame(df)
df['Submission_Date'] = pd.to_datetime(df['Submission_Date'])
display(df)

这是一个示例实现：

unique_flairs = list(df['Flair'].unique())
flair_df = pd.DataFrame()

# iterate over unique dates
for date in df['Submission_Date'].unique():
    date_subset = df.loc[df['Submission_Date'] == date]
    # for flairs in each date, get counts of values
    counts = date_subset['Flair'].value_counts()
    # get shares
    shares = counts / len(date_subset)
    # transpose series for appending
    shares_df = pd.DataFrame(shares).transpose()
    shares_df['Submission_Date'] = date
    for flair in [x for x in unique_flairs if x not in shares_df.columns]:
        shares_df[flair] = np.nan
    # append shares per date
    flair_df = pd.concat([flair_df, shares_df])
    flair_df = flair_df.reset_index(drop=True)

# rearrange columns
flair_df = flair_df[['Submission_Date'] + unique_flairs]
display(flair_df)

输出：

Answer 2

您可以使用数据透视表来实现此结果。我用一些随机值组成了输入数据框。

data = pandas.DataFrame.from_dict({
    "Submission_Date": [
        datetime.date(2021, 1, 1),
        datetime.date(2021, 1, 1),
        datetime.date(2021, 1, 2),
        datetime.date(2021, 1, 2),
        datetime.date(2021, 1, 3),
        datetime.date(2021, 1, 3),
        datetime.date(2021, 1, 3),
        datetime.date(2021, 1, 4),
    ],
    "Flair": ["Discussion", "Due Diligence", "Due Diligence", "Discussion", "Discussion",  "Hedge Fund Tears", "News", "News"],
})
data["Flair1"] = data.Flair.values # copy to another column to assist pivot
res = pandas.pivot_table(
    data,
    index=["Submission_Date"],
    values=['Flair'],
    columns=['Flair1'],
    aggfunc='count',
    fill_value=0

)
res = pandas.DataFrame(res.to_records())
res.columns = [col.replace("('Flair', ", '').replace(")", '') for col in res.columns]
res['Total'] = res.astype({col:float for col in res.columns if col != "Submission_Date"}).sum(numeric_only=True, axis=1) # find Total
res[[col for col in res.columns if col != "Submission_Date"]] = res[[col for col in res.columns if col != "Submission_Date"]].div(res.Total, axis=0) # divide by Total
res = res.drop(columns=['Total']) # drop Total
print(res)