import pandas as pd import numpy as np import matplotlib.pyplot as plt dates = pd.date_range(‘20180116‘, periods=3) # 创建 16 17 18 等六个日期 df = pd.DataFrame(np.random.randn(3,4), index=dates, columns=list(‘ABCD‘)) # 这是二维的，类似于一个 # Getting # print(df[‘A‘]) # 选中一列 # 2013-01-01 0.469112 # 2013-01-02 1.212112 # 2013-01-03 -0.861849 # 2013-01-04 0.721555 # 2013-01-05 -0.424972 # 2013-01-06 -0.673690 # Freq: D, Name: A, dtype: float64 # print(df[0:3]) # 不包括第三行！ # A B C D # 2018-01-16 -0.621070 -0.558260 -0.068434 -1.225484 # 2018-01-17 0.500783 -0.289074 -0.251468 -0.935832 # 2018-01-18 0.299410 2.279664 0.325912 0.461620 # print(df[‘20180116‘:‘20180117‘]) # 顾名思义 # A B C D # 2018-01-16 -0.009937 0.545212 0.682592 0.666001 # 2018-01-17 0.641140 0.539408 0.876006 -0.410707 # Selection by Label # print(df) # print(df.loc[dates[0]]) # A B C D # 2018-01-16 -1.531173 0.473732 -0.017051 -0.911358 # 2018-01-17 -2.153974 1.320710 1.970252 -1.992209 # 2018-01-18 -0.829090 1.096573 0.997688 -0.401185 # A -1.531173 # B 0.473732 # C -0.017051 # D -0.911358 # Name: 2018-01-16 00:00:00, dtype: float64 # print(df) # print(df.loc[:,[‘A‘,‘B‘]]) # A B C D # 2018-01-16 0.077497 1.364726 0.343679 -1.099019 # 2018-01-17 -0.573355 -0.939503 0.020275 1.073868 # 2018-01-18 -0.507676 -0.820279 -1.802128 -0.328978 # A B # 2018-01-16 0.077497 1.364726 # 2018-01-17 -0.573355 -0.939503 # 2018-01-18 -0.507676 -0.820279 # print(df) # print(df.loc[‘20180116‘:‘20180117‘,[‘A‘,‘B‘]]) # A B C D # 2018-01-16 2.526965 0.820404 0.095466 0.611306 # 2018-01-17 -1.359352 1.602012 0.337596 2.380324 # 2018-01-18 -0.453608 1.454857 1.443562 2.145979 # A B # 2018-01-16 2.526965 0.820404 # 2018-01-17 -1.359352 1.602012 # print(df) # print(df.loc[‘20180116‘,[‘A‘,‘B‘]]) # A B C D # 2018-01-16 -0.143268 -0.954798 0.637066 -1.433980 # 2018-01-17 0.527822 1.673820 1.150244 -0.644368 # 2018-01-18 0.550647 0.012898 1.065985 2.614110 # A -0.143268 # B -0.954798 # Name: 2018-01-16 00:00:00, dtype: float64 # print(df) # print(df.loc[dates[0],‘A‘]) # A B C D # 2018-01-16 0.557596 -0.140733 0.921194 -0.618365 # 2018-01-17 0.499742 -0.709669 -0.128449 -3.033026 # 2018-01-18 0.014871 -1.198496 -0.241682 -0.502687 # 0.5575964215814226 # print(df) # print(df.at[dates[0],‘A‘]) # at的使用方法与loc类似，但是比loc有更快的访问数据的速度，而且只能访问单个元素，不能访问多个元素。 # A B C D # 2018-01-16 0.557596 -0.140733 0.921194 -0.618365 # 2018-01-17 0.499742 -0.709669 -0.128449 -3.033026 # 2018-01-18 0.014871 -1.198496 -0.241682 -0.502687 # 0.5575964215814226 # Selection by Position # print(df) # print(df.iloc[0]) # print(df.iloc[2]) # A B C D # 2018-01-16 -0.660315 0.116266 -0.914127 0.598307 # 2018-01-17 -1.882812 1.715777 -0.355752 -0.192475 # 2018-01-18 0.628092 0.700135 0.402080 0.949126 # A -0.660315 # B 0.116266 # C -0.914127 # D 0.598307 # Name: 2018-01-16 00:00:00, dtype: float64 # A 0.628092 # B 0.700135 # C 0.402080 # D 0.949126 # Name: 2018-01-18 00:00:00, dtype: float64 # print(df) # print(df.iloc[0:1,1:3]) # [0:1] 不包括 1 ， [1:3] 不包括 3 # A B C D # 2018-01-16 -0.685245 1.835675 -0.630813 -0.408195 # 2018-01-17 -0.899057 0.257409 0.305275 -0.956311 # 2018-01-18 -1.111117 0.280925 -0.463713 0.882284 # B C # 2018-01-16 1.835675 -0.630813 # print(df) # print(df.iloc[[1,2,0],[0,2]]) # 选第2行、第3行、第0行，第1列第3列 # print(df.iloc[1:2,:]) # print(df.iloc[:,1:2]) # A B C D # 2018-01-16 0.221714 0.357890 -0.905870 -0.099446 # 2018-01-17 -0.636384 -1.428893 -0.471488 -1.197841 # 2018-01-18 1.044619 -0.346529 -0.164955 0.201145 # A C # 2018-01-17 -0.636384 -0.471488 # 2018-01-18 1.044619 -0.164955 # 2018-01-16 0.221714 -0.905870 # A B C D # 2018-01-17 -0.636384 -1.428893 -0.471488 -1.197841 # B # 2018-01-16 0.357890 # 2018-01-17 -1.428893 # 2018-01-18 -0.346529 # print(df.iloc[1,1]) # print(df.iat[1,1]) # 访问确切的值 比上面的快？ # -0.2891820477026986 # -0.2891820477026986 # Boolean Indexing # print(df[df.A > 0]) # 多随机几次是有可能 empty set 的，选中的就是 df.A > 0 的那些行！ # A B C D # 2018-01-17 0.322452 0.803659 -0.982818 0.149446 # 2018-01-18 0.501591 -0.114393 -0.306871 -2.258557 # 上面几列都是 A 列数字大于 0 的 # print(df[df > 0]) # 这个是全局选值 # A B C D # 2018-01-16 1.453356 NaN 0.120802 0.368208 # 2018-01-17 0.459706 0.802484 NaN NaN # 2018-01-18 NaN 0.569428 0.952326 0.541748 # Setting # Setting a new column automatically aligns the data by the indexes # s1 = pd.Series([1, 2, 3], index=pd.date_range(‘20180116‘, periods=3)) # print(s1) # print(df) # df[‘F‘] = s1 # print(df) # # 2018-01-16 1 # 2018-01-17 2 # 2018-01-18 3 # Freq: D, dtype: int64 # A B C D # 2018-01-16 -0.261046 -0.561609 -2.263514 2.359545 # 2018-01-17 0.563822 -1.301185 0.906939 0.478209 # 2018-01-18 0.942304 1.231033 -0.016457 0.659738 # A B C D F # 2018-01-16 -0.261046 -0.561609 -2.263514 2.359545 1 # 2018-01-17 0.563822 -1.301185 0.906939 0.478209 2 # 2018-01-18 0.942304 1.231033 -0.016457 0.659738 3 # print(df) # df.at[dates[0],‘A‘] = 0 # Setting values by label # df.iat[0, 1] = 0 # Setting values by position # df.loc[:,‘D‘] = np.array([99] * len(df)) # Setting by assigning with a numpy array # print(df) # A B C D # 2018-01-16 1.113651 -0.978514 -0.852811 0.933365 # 2018-01-17 -1.395547 -0.158742 -1.509723 -0.917854 # 2018-01-18 0.672396 -1.248654 -1.430043 -1.133012 # A B C D # 2018-01-16 0.000000 0.000000 -0.852811 99 # 2018-01-17 -1.395547 -0.158742 -1.509723 99 # 2018-01-18 0.672396 -1.248654 -1.430043 99 # A where operation with setting. # df2 = df.copy() # print(df2) # df2[df2 > 0] = -df2 # print(df2) # A B C D # 2018-01-16 0.824635 -0.914218 -0.953014 0.166094 # 2018-01-17 -0.037925 0.018838 0.927026 0.322848 # 2018-01-18 0.596024 0.851863 -0.548556 0.243168 # A B C D # 2018-01-16 -0.824635 -0.914218 -0.953014 -0.166094 # 2018-01-17 -0.037925 -0.018838 -0.927026 -0.322848 # 2018-01-18 -0.596024 -0.851863 -0.548556 -0.243168