カテゴリデータの変換と分析

Pythonによる探索的データ分析

George Boorman

Curriculum Manager, DataCamp

データをプレビュー

print(salaries.select_dtypes("object").head())

  Designation                 Experience    Employment_Status    Employee_Location    Company_Size
0 Data Scientist              Mid           FT                   DE                   L
1 Machine Learning Scientist  Senior        FT                   JP                   S
2 Big Data Engineer           Senior        FT                   GB                   M
3 Product Data Analyst        Mid           FT                   HN                   S
4 Machine Learning Engineer   Senior        FT                   US                   L

職種名

print(salaries["Designation"].value_counts())

Data Scientist                              143
Data Engineer                               132
Data Analyst                                 97
Machine Learning Engineer                    41
Research Scientist                           16
Data Science Manager                         12
Data Architect                               11
Big Data Engineer                             8
Machine Learning Scientist                    8
...

職種名

print(salaries["Designation"].nunique())

職種名

Bar plot displaying the five most popular data roles - data scientist, data engineer, data analyst, machine learning engineer, and research scientist

カテゴリから有用な情報を引き出す

現在の形式では、洞察を得にくい
pandas.Series.str.contains()
- 列内で特定の文字列や複数の文字列を検索

salaries["Designation"].str.contains("Scientist")

0       True
1       True
2      False
3      False
       ...  
604    False
605    False
606     True
Name: Designation, Length: 607, dtype: bool

文字列内で複数の語句を検索

対象とする語句： Machine Learning または AI

salaries["Designation"].str.contains("Machine Learning|AI")

0      False
1       True
2      False
3      False
       ...  
604    False
605    False
606     True
Name: Designation, Length: 607, dtype: bool

文字列内で複数の語句を検索

対象とする語句： Dataで始まるもの

salaries["Designation"].str.contains("^Data")

0       True
1      False
2      False
3      False
       ...  
604     True
605     True
606    False
Name: Designation, Length: 607, dtype: bool

文字列内で複数の語句を検索

job_categories = ["Data Science", "Data Analytics", 
                  "Data Engineering", "Machine Learning",
                  "Managerial", "Consultant"]

文字列内で複数の語句を検索

data_science = "Data Scientist|NLP"

data_analyst = "Analyst|Analytics"

data_engineer = "Data Engineer|ETL|Architect|Infrastructure"

ml_engineer = "Machine Learning|ML|Big Data|AI"

manager = "Manager|Head|Director|Lead|Principal|Staff"

consultant = "Consultant|Freelance"

文字列内で複数の語句を検索

conditions = [
    (salaries["Designation"].str.contains(data_science)),

    (salaries["Designation"].str.contains(data_analyst)),

    (salaries["Designation"].str.contains(data_engineer)),

    (salaries["Designation"].str.contains(ml_engineer)),

    (salaries["Designation"].str.contains(manager)),

    (salaries["Designation"].str.contains(consultant))
]

カテゴリ列の作成

salaries["Job_Category"] =

カテゴリ列の作成

salaries["Job_Category"] = np.select(conditions,

カテゴリ列の作成

salaries["Job_Category"] = np.select(conditions, 
                                     job_categories,

カテゴリ列の作成

salaries["Job_Category"] = np.select(conditions, 
                                     job_categories, 
                                     default="Other")

職種カテゴリをプレビュー

print(salaries[["Designation", "Job_Category"]].head())

                  Designation      Job_Category
0              Data Scientist      Data Science
1  Machine Learning Scientist  Machine Learning
2           Big Data Engineer  Data Engineering
3        Product Data Analyst    Data Analytics
4   Machine Learning Engineer  Machine Learning

職種カテゴリの頻度を可視化

sns.countplot(data=salaries, x="Job_Category")

plt.show()

Bar plot displaying the count of jobs by category

お疲れさまでした！

Pythonによる探索的データ分析