pandas数据类型之dataframe

Pandas 简介：

python数据分析library
是基于Numpy构建的一个library
有一种用python做Excel/SQL/R的感觉

现在流行的机器学习框架 Tensorflow/PyTorch 等等，语法都与Numpy比较接近

# encoding=utf-8

import numpy as np
import pandas as pd


def main():
    """
    一个Dataframe就是一张表格，Series表示的是一维数组，Dataframe则是一个二维数组，
    可以类比成一张excel的spreadsheet，也可以理解成数据库中的一张表，
    最后也可以把Dataframe当做一组Series的集合
    """
    # 1.创建一个Dataframe， dataframe可以由 dictionary or list 构建得到
    data = {
        "city": ["beijing", "shanghai", "guangzhou", "shenzhen", "wuhan"],
        "year": [2016, 2017, 2016, 2017, 2016],
        "population": [2100, 2300, 1000, 700, 500]
    }
    # columns的名字和顺序可以指定，如果指定不存在的column，会生成对应的column并设置为None
    data_frame = pd.DataFrame(data, columns=["city", "year", "population", "debt"])
    print(data_frame, "\n", type(data_frame))
    """
            city  year  population  debt
    0    beijing  2016        2100  NaN
    1   shanghai  2017        2300  NaN
    2  guangzhou  2016        1000  NaN
    3   shenzhen  2017         700  NaN
    4      wuhan  2016         500  NaN 
    <class 'pandas.core.frame.DataFrame'>
    """
    # 可以指定index的名称顺序，如果有多余的index会报错！
    data_frame_index = pd.DataFrame(data, index=["one", "two", "three", "four", "five"])
    print(data_frame_index, "\n", type(data_frame_index))
    """
                city  population  year
    one      beijing        2100  2016
    two     shanghai        2300  2017
    three  guangzhou        1000  2016
    four    shenzhen         700  2017
    five       wuhan         500  2016 
    <class 'pandas.core.frame.DataFrame'>
    """

    # 2.从Dataframe中选择数据
    city_series = data_frame["city"]
    print(city_series, "\n", type(city_series))  # 得到的是一个series
    """
    0      beijing
    1     shanghai
    2    guangzhou
    3     shenzhen
    4        wuhan
    Name: city, dtype: object 
    <class 'pandas.core.series.Series'>
    """
    # 上面的data_frame["city"]也可以写成dataframe.city
    print(data_frame.city)
    # 怎样读一行的数据呢？
    one_series = data_frame.ix[1]
    print(one_series, "\n", type(one_series))
    """
    city          shanghai
    year              2017
    population        2300
    debt               NaN
    Name: 1, dtype: object 
    <class 'pandas.core.series.Series'>
    """
    # 上面的data_frame.ix[1]不能写成data_frame.ix.1
    # 可以写成data_frame.ix[index_name]
    # 切片取值，是根据index来取
    print(data_frame[2:-1])  # 根据默认的index取值是顾前不顾尾
    """
            city  year  population debt
    2  guangzhou  2016        1000  NaN
    3   shenzhen  2017         700  NaN
    """
    # 根据手动指定的index来切片
    print(data_frame_index["one": "three"])  # 使用指定的index取值是顾前又顾尾
    """
                city  population  year
    one      beijing        2100  2016
    two     shanghai        2300  2017
    three  guangzhou        1000  2016
    """
    # 同时指定 index and columns 来切片，切完之后发现不像切片，像坐标点取值["one", "city"] ["one", "year"] ...
    print(data_frame_index.ix[["one", "three"], ["city", "year"]])
    """
                city  year
    one      beijing  2016
    three  guangzhou  2016
    """
    # 这个像范围指定，适用于对列的切片，例如：[:, "city": "year"]
    print(data_frame_index.ix["one": "three", "city": "year"])
    """
                city  population  year
    one      beijing        2100  2016
    two     shanghai        2300  2017
    three  guangzhou        1000  2016    
    """
    # 使用 boolean indexing 来过滤行
    print(data_frame_index[data_frame_index["population"] > 2000])
    """
             city  population  year
    one   beijing        2100  2016
    two  shanghai        2300  2017
    """
    print(data_frame_index[data_frame_index.index == "one"])
    """
            city  population  year
    one  beijing        2100  2016
    """
    # 取columns的内容
    print(data_frame.columns)
    print(type(data_frame.columns))
    """
    Index(['city', 'year', 'population', 'debt'], dtype='object')
    <class 'pandas.core.indexes.base.Index'>
    """
    # 取index的内容
    print(data_frame.index)
    print(type(data_frame.index))
    """
    RangeIndex(start=0, stop=5, step=1)
    <class 'pandas.core.indexes.range.RangeIndex'>
    """

    # 3.给Dataframe赋值
    # 给单个数据赋值
    data_frame["population"][0] = 2222
    # data_frame.ix[4]["year"] = 2019  这样貌似用不了
    # 给整列赋值
    data_frame["debt"] = -200000
    # 给整行赋值
    data_frame.ix[4] = 0
    print(data_frame)
    """
            city  year  population    debt
    0    beijing  2016        2222 -200000
    1   shanghai  2017        2300 -200000
    2  guangzhou  2016        1000 -200000
    3   shenzhen  2017         700 -200000
    4          0     0           0       0
    """
    # 还可以把一个ndarray赋值给某一行、某一列
    data_frame["debt"] = np.arange(0, 5)
    data_frame.ix[4] = np.arange(1, 5)
    print(data_frame)
    """
            city  year  population  debt
    0    beijing  2016        2222     0
    1   shanghai  2017        2300     1
    2  guangzhou  2016        1000     2
    3   shenzhen  2017         700     3
    4          1     2           3     4
    """
    # 还可以用Series来指定需要修改的index以及相对应的value，没有指定的默认用None
    value_series = pd.Series([100, 200, 300])  # 在指定的index名称的情况下也适用
    data_frame.debt = value_series  # 将一个series赋值给data_frame
    print(data_frame)
    """
            city  year  population   debt
    0    beijing  2016        2222  100.0
    1   shanghai  2017        2300  200.0
    2  guangzhou  2016        1000  300.0
    3   shenzhen  2017         700    NaN
    4          1     2           3    NaN
    """
    # 使用直接赋值的方式增加一个column
    data_frame["western"] = (data_frame.city == "beijing")
    # data_frame.western = (data_frame.city == "beijing")  # 不存在western列的情况下不能这么写
    print(data_frame)
    """
            city  year  population   debt  western
    0    beijing  2016        2222  100.0     True
    1   shanghai  2017        2300  200.0    False
    2  guangzhou  2016        1000  300.0    False
    3   shenzhen  2017         700    NaN    False
    4          1     2           3    NaN    False
    """

    # 4.Dataframe转置 和 数据的重新组织
    print(data_frame.T)
    """
                      0         1          2         3      4
    city        beijing  shanghai  guangzhou  shenzhen      1
    year           2016      2017       2016      2017      2
    population     2222      2300       1000       700      3
    debt            100       200        300       NaN    NaN
    western        True     False      False     False  False
    """
    # 数据的重新组织
    populations = {
        "beijing": {2016: 2100, 2017: 2200},
        "shanghai": {2015: 2400, 2016: 2500, 2017: 2600}
    }
    pop_data_frame = pd.DataFrame(populations)
    print(pop_data_frame)
    """
          beijing  shanghai
    2015      NaN      2400
    2016   2100.0      2500
    2017   2200.0      2600
    """
    print(pop_data_frame.T)
    """   
                2015    2016    2017
    beijing      NaN  2100.0  2200.0
    shanghai  2400.0  2500.0  2600.0
    """
    # 获取两个城市的人口数据的series
    beijing_series = pop_data_frame["beijing"][:-1]   # 不取最后一个
    print(beijing_series)
    """
    2015       NaN
    2016    2100.0
    Name: beijing, dtype: float64
    """
    shanghai_series = pop_data_frame["shanghai"][:-1]
    print(shanghai_series)
    """
    2015    2400
    2016    2500
    Name: shanghai, dtype: int64
    """
    # 重新构建data_frame
    pdata = {"beijing": beijing_series, "shanghai": shanghai_series}
    pdata_frame = pd.DataFrame(pdata)
    print(pdata_frame)
    """
          beijing  shanghai
    2015      NaN      2400
    2016   2100.0      2500
    """

    # 5.几个常用的function
    # 指定index and columns 的名字
    pdata_frame.columns.name = "city"
    pdata_frame.index.name = "year"
    print(pdata_frame)
    """
    city  beijing  shanghai
    year                   
    2015      NaN      2400
    2016   2100.0      2500
    """
    # 取values
    print(pdata_frame["beijing"].values, "\n", type(pdata_frame.values))
    """
    [ nan 2100.] 
    <class 'numpy.ndarray'>
    """


if __name__ == '__main__':
    main()