如何在Pandas DataFrame中根据给定列中的计数只保留行

Question

我有一个Pandas DataFrame，其中一个列中有一些分类数据。在对特定专栏执行value_counts时，我得到类似于以下内容的内容：

HR                          176
Coding                       81
Reject                       74
Database Administration      21
Finance                      17
Project Management           16
Sales                        15
DevOps                       13
Core Electronics             10
Networking                   10
Medical Science               9
Core Mechanical               8
Web Development               4
Puzzles                       3
behavioural                   3
not a question                2
civil engineering             1
Mathematics                   1
Finance, Medical Science      1
Sales, HR                     1

我想做的是只保留一个计数>=的类别某个阈值(例如10)。所有较小的类别都应单独归入“其他”类别，即结果应该如下所示：

HR                          176
Coding                       81
Reject                       74

*Other*                      33

Database Administration      21
Finance                      17
Project Management           16
Sales                        15
DevOps                       13
Core Electronics             10
Networking                   10

我过去做过这样的工作，方法是将一个defaultdict(int)合并在一起，并且只使用计数>=阈值的实例。我想知道是否有一种Pandas规范的方法来实现同样的目标。

Answer 1

这就是你要找的答案吗？

Pandas: Selecting rows based on value counts of a particular column

否则也许这就是你想要的：

data = pd.DataFrame([["researcher",150],["politician",15],["builder",1],["teacher",5],])
data.columns = ["category", "count"]
filter_value = 10
d1 = data[data['count'] >= filter_value]
d2 = data[data['count'] < filter_value]
d1["tag"] = "filter_passed"
d2["tag"] = "Others"
data = pd.concat([d1,d2])
>>> data
     category  count            tag
0  researcher    150  filter_passed
1  politician     15  filter_passed
2     builder      1         Others
3     teacher      5         Others

Answer 2

我会使用一个掩码来执行布尔索引和concat

m = s>=10
out = (pd.concat([s[m], pd.Series(s[~m].sum(), index=['Others'])])
         .sort_values(ascending=False)
      )

产出：

HR                         176
Coding                      81
Reject                      74
Others                      33
Database Administration     21
Finance                     17
Project Management          16
Sales                       15
DevOps                      13
Core Electronics            10
Networking