pandas:填充缺失值 fillna("missing") 和fillna("missing",inplace=True)的區別
阿新 • • 發佈：2019-01-23
當資料中存在NaN缺失值時，我們可以用其他數值替代NaN，主要用到了DataFrame.fillna()方法，下面我們來看看具體的用法：
1.先來建立一個帶有缺失值的DataFrame:
# coding=utf-8 import numpy as np import xgboost as xgb from sklearn.linear_model import LogisticRegression from sklearn.model_selection import train_test_split from sklearn.ensemble import RandomForestRegressor from sklearn.ensemble import RandomForestClassifier import pandas as pd import csv from pandas import DataFrame df=pd.DataFrame(np.random.randn(5,3),index=list('abcde'),columns=['one','two','three']) print df #原有資料 print('\n') df.ix[1,:-1]=np.nan #第1行，從第0列到倒數第1列 的所有資料填充為NaN （都是從0行，0列開始的 ） df.ix[1:-1,2]=np.nan #第1行到倒數第1行，第2列 的所有資料填充為NaN （都是從0行，0列開始的 ） print df #被填充有空值的資料 print('\n') print df.fillna("missing") #此資料已被填充 print('\n') print df #df 的內容還是原來有空值的資料 print('\n') print df.fillna(method='pad') #用前一個數據代替NaN：method='pad' print('\n') print df.fillna(method='bfill',limit=1) #與pad相反，bfill表示用後一個數據代替NaN print('\n') print df.fillna(df.mean()) #使用平均數或者其他描述性統計量來代替NaN print('\n') print df.fillna(df.mean()['one':'two']) #選擇哪一列進行缺失值的處理
結果：
2、 fillna("missing") 和fillna("missing",inplace=True)的區別one two three a 0.348287 -0.579763 -0.687073 b -0.128967 1.734375 -1.530778 c 0.448428 -0.791999 0.620952 d 0.808736 -0.554402 -0.094709 e 1.553160 -1.336362 -0.159426 one two three a 0.348287 -0.579763 -0.687073 b NaN NaN NaN c 0.448428 -0.791999 NaN d 0.808736 -0.554402 NaN e 1.553160 -1.336362 -0.159426 one two three a 0.348287 -0.579763 -0.687073 b missing missing missing c 0.448428 -0.791999 missing d 0.808736 -0.554402 missing e 1.55316 -1.33636 -0.159426 one two three a 0.348287 -0.579763 -0.687073 b NaN NaN NaN c 0.448428 -0.791999 NaN d 0.808736 -0.554402 NaN e 1.553160 -1.336362 -0.159426 one two three a 0.348287 -0.579763 -0.687073 b 0.348287 -0.579763 -0.687073 c 0.448428 -0.791999 -0.687073 d 0.808736 -0.554402 -0.687073 e 1.553160 -1.336362 -0.159426 one two three a 0.348287 -0.579763 -0.687073 b 0.448428 -0.791999 NaN c 0.448428 -0.791999 NaN d 0.808736 -0.554402 -0.159426 e 1.553160 -1.336362 -0.159426 one two three a 0.348287 -0.579763 -0.687073 b 0.789653 -0.815631 -0.423249 c 0.448428 -0.791999 -0.423249 d 0.808736 -0.554402 -0.423249 e 1.553160 -1.336362 -0.159426 one two three a 0.348287 -0.579763 -0.687073 b 0.789653 -0.815631 NaN c 0.448428 -0.791999 NaN d 0.808736 -0.554402 NaN e 1.553160 -1.336362 -0.159426
# coding=utf-8 import numpy as np import xgboost as xgb from sklearn.linear_model import LogisticRegression from sklearn.model_selection import train_test_split from sklearn.ensemble import RandomForestRegressor from sklearn.ensemble import RandomForestClassifier import pandas as pd import csv from pandas import DataFrame df=pd.DataFrame(np.random.randn(5,3),index=list('abcde'),columns=['one','two','three']) print df #原有資料 print('\n') df.ix[1,:-1]=np.nan #第1行，從第0列到倒數第1列 的所有資料填充為NaN （都是從0行，0列開始的 ） df.ix[1:-1,2]=np.nan #第1行到倒數第1行，第2列 的所有資料填充為NaN （都是從0行，0列開始的 ） print df #被填充有空值的資料 print('\n') print df.fillna("missing") #df 原資料沒有被 missing 填充，df 的資料沒有變 print('\n') print df #df 原資料沒有被 missing 填充，df 的資料沒有變 print('\n') df.fillna("missing",inplace=True) #df 原資料已被 missing 填充 print df #df 原資料已被 missing 填充 print "----------"
結果：
one two three
a 0.428457 -0.797473 -0.448647
b -1.744598 -0.944395 0.952140
c 1.096071 0.812616 1.980379
d -1.120961 1.193119 0.455609
e 1.039164 -0.384459 0.289628
one two three
a 0.428457 -0.797473 -0.448647
b NaN NaN NaN
c 1.096071 0.812616 NaN
d -1.120961 1.193119 NaN
e 1.039164 -0.384459 0.289628
one two three
a 0.428457 -0.797473 -0.448647
b missing missing missing
c 1.09607 0.812616 missing
d -1.12096 1.19312 missing
e 1.03916 -0.384459 0.289628
one two three
a 0.428457 -0.797473 -0.448647
b NaN NaN NaN
c 1.096071 0.812616 NaN
d -1.120961 1.193119 NaN
e 1.039164 -0.384459 0.289628
one two three
a 0.428457 -0.797473 -0.448647
b missing missing missing
c 1.09607 0.812616 missing
d -1.12096 1.19312 missing
e 1.03916 -0.384459 0.289628
----------