为什么使用 qcut 后我的数据值被“NaN”替换了？答案

【问题标题】：Why have my data values been replaced by "NaN" after using qcut?为什么使用 qcut 后我的数据值被“NaN”替换了？
【发布时间】：2020-09-16 23:16:40
【问题描述】：

我正在使用 9000 行和 6 列的 pandas 数据框。在这一点上，我正在尝试将工作的连续变量“经验”年数转换为 4 个工作（商业经理 - 业务开发人员 -网络营销商 - 流量管理器）。

考虑到每个工作的经验范围不一样，我使用“qcut”将数据分为4组，如下所示：

（您可以运行下面的代码来获取数据框示例）

import pandas as pd


df = pd.DataFrame({'Job': ['Commercial Manager', 'Traffic Manager', 'Web Marketer', 'Commercial Manager', 'Commercial Manager', 'Web Marketer', 'Commercial Manager', 'Commercial Manager', 'Traffic Manager', 'Business Developer', 'Business Developer', 'Web Marketer', 'Traffic Manager', 'Traffic Manager', 'Commercial Manager', 'Business Developer', 'Traffic Manager', 'Commercial Manager', 'Business Developer', 'Business Developer', 'Web Marketer'], 
                   'Experience': [1.00000, 3.00000, 3.00000, 1.50000, 2.00000, 6.00000, 0.00000, 4.00000, 8.00000, 5.00000, 0.50000, 3.00000, 3.00000, 0.00000, 2.00000, 3.00000, 0.50000, 3.00000, 3.00000, 8.00000, 3.50000]})


levels = ["beginner", "intermediate", "advanced", "expert"]
jobs = ["Commercial Manager", "Business Developer", "Web Marketer", "Traffic Manager"]


def convert(levels, jobs):
  for j in jobs:
    df["Level"] = pd.qcut(df.loc[df["Job"] == j, "Experience"].rank(method="first"), q = 4, labels = levels, duplicates = "drop")
  return df

convert(levels, jobs)

这是使用“qcut”后的输出：

    Job                     Experience       Level 
0   Commercial Manager      1.00000          NaN
1   Traffic Manager         3.00000          intermediate
2   Web Marketer            3.00000          NaN
3   Commercial Manager      1.50000          NaN
4   Commercial Manager      2.00000          NaN
5   Web Marketer            6.00000          NaN
6   Commercial Manager      0.00000          NaN
7   Commercial Manager      4.00000          NaN
8   Traffic Manager         8.00000          expert
9   Business Developer      5.00000          NaN 
10  Business Developer      0.50000          NaN 
11  Web Marketer            3.00000          NaN 
12  Traffic Manager         3.00000          intermediate
13  Traffic Manager         0.00000          beginner
14  Commercial Manager      2.00000          NaN
15  Business Developer      3.00000          NaN
16  Traffic Manager         0.50000          beginner
17  Commercial Manager      3.00000          NaN
18  Business Developer      3.00000          NaN
19  Business Developer      8.00000          NaN
20  Web Marketer            3.50000          NaN

它似乎只适用于“流量管理器”，它用 NaN 取代了其他 level 体验。我真的迷路了。有什么帮助吗？

【问题讨论】：

其中一些解决方案可能会有所帮助：stackoverflow.com/a/40548606/5783745
谢谢，不过我已经看过了，没用。仍然卡住:(
你不需要那个循环

标签： python pandas dataframe binning

【解决方案1】：

您想在 groupby 操作中执行此操作：

import numpy
import pandas

levels = ["beginner", "intermediate", "advanced", "expert"]
jobs = ["Commercial Manager", "Business Developer", "Web Marketer", "Traffic Manager"]

df = pandas.DataFrame({
    'Job': numpy.random.choice(levels, size=150), 
    'Experience': numpy.random.uniform(0.25, 10.5, size=150)
}).assign(
    level=df.groupby(['Job'])['Experience'] # for each unique job...
            # apply a quantile (quartile) cut 
            .apply(lambda g: pd.qcut(g, q=4, labels=levels, duplicates="drop"))
)

【讨论】：

【解决方案2】：

  # I would just change two things to what Paul suggested (jobs instead of levels and the rank(method="first") because there was still an error:

levels = ["beginner", "intermediate", "advanced", "expert"]
jobs = ["Commercial Manager", "Business Developer", "Web Marketer", "Traffic Manager"]

df = pandas.DataFrame({
  'Job': numpy.random.choice(jobs, size=150), 
  'Experience': numpy.random.uniform(0.25, 10.5, size=150)
}).assign(
  level=df.groupby(['Job'])['Experience'] # for each unique job...
        # apply a quantile (quartile) cut 
        .apply(lambda g: pd.qcut(g.rank(method="first"), q=4, labels=levels, duplicates="drop"))
)

【讨论】：