Python中使用预设排名对字符串变量频率分布排序失效问题排查

Hey there! I see the issue with your code—let's break it down and fix that custom sorting problem.

What's Going Wrong #

Your line fd_education = fd_education.sort_index(level='education_rank') isn't doing what you expect. The level parameter in sort_index() is designed for multi-level indexes (when your index has multiple hierarchical levels), but your fd_education has a plain string index of education categories. So this line essentially does nothing, and sort_index() falls back to its default behavior: alphabetical sorting.

Two Working Solutions #

Solution 1: Sort Using Your Custom Rank Map Directly

We can sort the index of your frequency series using the rank values from your education_rank dictionary, then reorder the series with that sorted index:

import pandas as pd

# Your custom rank dictionary
education_rank = {' Bachelors':12, ' HS-grad':8, ' 11th':6, ' Masters':14, ' 9th':5, ' Some-college':11, ' Assoc-acdm':10, ' Assoc-voc':9, ' 7th-8th':4, ' Doctorate':15, ' Prof-school':13, ' 5th-6th':3, ' 10th':16, ' 1st-4th':2, ' Preschool':1, ' 12th':7}

# Calculate frequency counts
fd_education = pd.value_counts(adult_data.education)

# Sort the index by your custom rank, then reorder the series
fd_education_sorted = fd_education.loc[sorted(fd_education.index, key=lambda x: education_rank[x])]

print(fd_education_sorted)

Here, sorted(fd_education.index, key=lambda x: education_rank[x]) sorts the education categories based on their corresponding rank values, and loc[] reorders the frequency series to match this sorted list.

Solution 2: Convert to Categorical Data (More Robust)

If you plan to work with this sorted order repeatedly, converting the education column to an ordered categorical type is a better approach. This way, any subsequent operations (like value_counts) will respect your custom order automatically:

import pandas as pd

# Your custom rank dictionary
education_rank = {' Bachelors':12, ' HS-grad':8, ' 11th':6, ' Masters':14, ' 9th':5, ' Some-college':11, ' Assoc-acdm':10, ' Assoc-voc':9, ' 7th-8th':4, ' Doctorate':15, ' Prof-school':13, ' 5th-6th':3, ' 10th':16, ' 1st-4th':2, ' Preschool':1, ' 12th':7}

# Get the education categories sorted by your rank (ascending order)
sorted_education_order = sorted(education_rank.keys(), key=lambda x: education_rank[x])

# Convert the education column to an ordered categorical
adult_data['education'] = pd.Categorical(
    adult_data['education'],
    categories=sorted_education_order,
    ordered=True
)

# Now value_counts will output in your custom order
fd_education = pd.value_counts(adult_data.education)
print(fd_education)

This method makes the custom order part of the column's metadata, so you don't have to re-sort every time you calculate frequencies.

内容的提问来源于stack exchange，提问作者Dat