Python数据科学编程-2.panda基本操作

df.reindex([1990, 1991, 1992,2001, 2002, 2003], columns=['Chengdu', 'Shenzhen', 'Beijing', 'Lanzhou'])。

panda基本操作

• Series
  • 创建
    • obj2 = pd.Series(np.random.randn(5), index = ['a', 'b', 'c', 'd', 'end'])
  • 设置索引
    • In [26]: dd = {'Beijing': 123, 'Shanghai': 456, 'Lanzhou': 789, 'Shenzhen': 1011} In [27]: pops = pd.Series(dd, index=['Beijing', 'Shanghai', 'Lanzhou', 'Hangzhou'])
  • 查看索引和值
    • obj.index
    • obj.values
  • 重设索引：Indexes can be re-assigned
    • pops.index = ['Recife', 'Salvador', 'Rio de Janeiro', 'Brasilia']
  • 设置name：Series and their indexes can have names
  • 分片：Slicing works, but it's peculiar with int indexes
    • pops.loc["Shanghai":]
    • In [124]: s = pd.Series(['a', 'b', 'c', 'd', 'e'], index=[1,2,3,67,9]) In [125]: s.loc[1:9]
• DF
  • 创建DF
    • 创建：Creating a DataFrame from a dictionary
      • frame = pd.DataFrame(data, columns=['year', 'pop', 'city'])
    • Creating a DataFrame from a dictionary of dictionaries
      • theDict = {'Beijing': {2001: 2.4, 2002: 3.1}, 'Lanzhou':{2002:5., 2003:1.7}}
  • 重设索引
    • df.reindex([1990, 1991, 1992,2001, 2002, 2003], columns=['Chengdu', 'Shenzhen', 'Beijing', 'Lanzhou'])
  • 删除行/列
    • frame.drop(['Japan', 'Argentina'],axis)
  • 转置：DataFrames can also be tranposed
  • 索引小操作：Some basic operations with Indexes
    • df.index.union(df.columns) df.index.intersection(df.columns)
• 进阶操作
  • 整行/列操作：The incredibly useful apply() method
    • df2.apply(lambda x: x.max() + x.min(), axis=1)
  • 排序：Sorting the index
    • df1.sort_index(ascending=False, axis=1)
  • 排行：Ranking values
    • df1.rank()
  • Uniqueness of Series elements
    • anotherSeries.isin(mySeries)
  • 合并Series
    • Adding up two DataFrames with add()
      • ser + ser2
      • ser.add(ser2, fill_value=0)
    • Adding a DataFrame and a Series
      • The Series is broadcasted to all the DataFrame columns
  • 筛选
    • year_sums[year_sums == year_sums.max()]
  • unique和（⭐）value-count
    • mySeries.unique() mySeries.value_counts()
• 索引并不copy：Indexing produces visualisations, not copies

panda基本操作

• Series
  • 创建
    • obj2 = pd.Series(np.random.randn(5), index = ['a', 'b', 'c', 'd', 'end'])
  • 设置索引
    • In [26]: dd = {'Beijing': 123, 'Shanghai': 456, 'Lanzhou': 789, 'Shenzhen': 1011} In [27]: pops = pd.Series(dd, index=['Beijing', 'Shanghai', 'Lanzhou', 'Hangzhou'])
  • 查看索引和值
    • obj.index
    • obj.values
  • 重设索引：Indexes can be re-assigned
    • pops.index = ['Recife', 'Salvador', 'Rio de Janeiro', 'Brasilia']
  • 设置name：Series and their indexes can have names
  • 分片：Slicing works, but it's peculiar with int indexes
    • pops.loc["Shanghai":]
    • In [124]: s = pd.Series(['a', 'b', 'c', 'd', 'e'], index=[1,2,3,67,9]) In [125]: s.loc[1:9]
• DF
  • 创建DF
    • 创建：Creating a DataFrame from a dictionary
      • frame = pd.DataFrame(data, columns=['year', 'pop', 'city'])
    • Creating a DataFrame from a dictionary of dictionaries
      • theDict = {'Beijing': {2001: 2.4, 2002: 3.1}, 'Lanzhou':{2002:5., 2003:1.7}}
  • 重设索引
    • df.reindex([1990, 1991, 1992,2001, 2002, 2003], columns=['Chengdu', 'Shenzhen', 'Beijing', 'Lanzhou'])
  • 删除行/列
    • frame.drop(['Japan', 'Argentina'],axis)
  • 转置：DataFrames can also be tranposed
  • 索引小操作：Some basic operations with Indexes
    • df.index.union(df.columns) df.index.intersection(df.columns)
• 进阶操作
  • 整行/列操作：The incredibly useful apply() method
    • df2.apply(lambda x: x.max() + x.min(), axis=1)
  • 排序：Sorting the index
    • df1.sort_index(ascending=False, axis=1)
  • 排行：Ranking values
    • df1.rank()
  • Uniqueness of Series elements
    • anotherSeries.isin(mySeries)
  • 合并Series
    • Adding up two DataFrames with add()
      • ser + ser2
      • ser.add(ser2, fill_value=0)
    • Adding a DataFrame and a Series
      • The Series is broadcasted to all the DataFrame columns
  • 筛选
    • year_sums[year_sums == year_sums.max()]
  • unique和（⭐）value-count
    • mySeries.unique() mySeries.value_counts()
• 索引并不copy：Indexing produces visualisations, not copies