最近用到Pandas，想查幾個(gè)函數(shù)，結(jié)果翻來翻去都是那幾個(gè)豆腐干大小的圖片，我真是緩緩地打出一個(gè)？難道你能對(duì)著一張圖片ctrl + F5？
于是趕緊安排一下官方庫(kù)的翻譯。這不，老鐵，來了~
本文翻譯整理自Pandas Cheat Sheet - Python for DataScience
點(diǎn)這里一鍵fork >>Pandas基礎(chǔ)命令速查清單
速查表內(nèi)容概要
[縮寫解釋 & 庫(kù)的導(dǎo)入]
[數(shù)據(jù)的導(dǎo)入]
[數(shù)據(jù)的導(dǎo)出]
[創(chuàng)建測(cè)試對(duì)象]
[數(shù)據(jù)的查看與檢查]
[數(shù)據(jù)的選取]
[數(shù)據(jù)的清洗]
[數(shù)據(jù)的過濾(filter),排序(sort)和分組(groupby)]
[數(shù)據(jù)的連接(join)與組合(combine)]
[數(shù)據(jù)的統(tǒng)計(jì)]
縮寫解釋 & 庫(kù)的導(dǎo)入
df --- 任意的pandas DataFrame(數(shù)據(jù)框)對(duì)象
s --- 任意的pandas Series(數(shù)組)對(duì)象
pandas和numpy是用Python做數(shù)據(jù)分析最基礎(chǔ)且最核心的庫(kù)
import pandas as pd # 導(dǎo)入pandas庫(kù)并簡(jiǎn)寫為pd
import numpy as np # 導(dǎo)入numpy庫(kù)并簡(jiǎn)寫為np
數(shù)據(jù)的導(dǎo)入
pd.read_csv(filename) # 導(dǎo)入csv格式文件中的數(shù)據(jù)
pd.read_table(filename) # 導(dǎo)入有分隔符的文本 (如TSV) 中的數(shù)據(jù)
pd.read_excel(filename) # 導(dǎo)入Excel格式文件中的數(shù)據(jù)
pd.read_sql(query, connection_object) # 導(dǎo)入SQL數(shù)據(jù)表/數(shù)據(jù)庫(kù)中的數(shù)據(jù)
pd.read_json(json_string) # 導(dǎo)入JSON格式的字符，URL地址或者文件中的數(shù)據(jù)
pd.read_html(url) # 導(dǎo)入經(jīng)過解析的URL地址中包含的數(shù)據(jù)框 (DataFrame) 數(shù)據(jù)
pd.read_clipboard() # 導(dǎo)入系統(tǒng)粘貼板里面的數(shù)據(jù)
pd.DataFrame(dict) # 導(dǎo)入Python字典 (dict) 里面的數(shù)據(jù)，其中key是數(shù)據(jù)框的表頭，value是數(shù)據(jù)框的內(nèi)容。
數(shù)據(jù)的導(dǎo)出
df.to_csv(filename) # 將數(shù)據(jù)框 (DataFrame)中的數(shù)據(jù)導(dǎo)入csv格式的文件中
df.to_excel(filename) # 將數(shù)據(jù)框 (DataFrame)中的數(shù)據(jù)導(dǎo)入Excel格式的文件中
df.to_sql(table_name,connection_object) # 將數(shù)據(jù)框 (DataFrame)中的數(shù)據(jù)導(dǎo)入SQL數(shù)據(jù)表/數(shù)據(jù)庫(kù)中
df.to_json(filename) # 將數(shù)據(jù)框 (DataFrame)中的數(shù)據(jù)導(dǎo)入JSON格式的文件中
創(chuàng)建測(cè)試對(duì)象
pd.DataFrame(np.random.rand(10,5)) # 創(chuàng)建一個(gè)5列10行的由隨機(jī)浮點(diǎn)數(shù)組成的數(shù)據(jù)框
DataFrame
pd.DataFrame(np.random.rand(10,5))
pd.Series(my_list) # 從一個(gè)可迭代的對(duì)象 my_list 中創(chuàng)建一個(gè)數(shù)據(jù)組
my_list = ['Kesci',100,'歡迎來到科賽網(wǎng)']
pd.Series(my_list)
df.index = pd.date_range('2017/1/1',periods=df.shape[0]) # 添加一個(gè)日期索引 index
df = pd.DataFrame(np.random.rand(10,5))
df.index = pd.date_range('2017/1/1',periods=df.shape[0])
df
數(shù)據(jù)的查看與檢查
df.head(n) # 查看數(shù)據(jù)框的前n行
df = pd.DataFrame(np.random.rand(10,5))
df.head(3)
df.tail(n) # 查看數(shù)據(jù)框的最后n行
df = pd.DataFrame(np.random.rand(10,5))
df.tail(3)
df.shape # 查看數(shù)據(jù)框的行數(shù)與列數(shù)
df = pd.DataFrame(np.random.rand(10,5))
df.shape
df.info() # 查看數(shù)據(jù)框(DataFrame) 的索引、數(shù)據(jù)類型及內(nèi)存信息
df = pd.DataFrame(np.random.rand(10,5))
df.info()
df.describe() # 對(duì)于數(shù)據(jù)類型為數(shù)值型的列，查詢其描述性統(tǒng)計(jì)的內(nèi)容
df.describe()
s.value_counts(dropna=False) # 查詢每個(gè)獨(dú)特?cái)?shù)據(jù)值出現(xiàn)次數(shù)統(tǒng)計(jì)
s = pd.Series([1,2,3,3,4,np.nan,5,5,5,6,7])
s.value_counts(dropna=False)
df.apply(pd.Series.value_counts) # 查詢數(shù)據(jù)框 (Data Frame) 中每個(gè)列的獨(dú)特?cái)?shù)據(jù)值出現(xiàn)次數(shù)統(tǒng)計(jì)
數(shù)據(jù)的選取
df[col] # 以數(shù)組Series 的形式返回選取的列
df =pd.DataFrame(np.random.rand(5,5),columns=list('ABCDE'))
df['C']
df[[col1, col2]] # 以新的數(shù)據(jù)框(DataFrame)的形式返回選取的列
df =pd.DataFrame(np.random.rand(5,5),columns=list('ABCDE'))
df[['B','E']]
s.iloc[0] # 按照位置選取
s =pd.Series(np.array(['I','Love','Data']))
s.iloc[0]
s.loc['index_one'] # 按照索引選取
s =pd.Series(np.array(['I','Love','Data']))
s.loc[1]
df.iloc[0,:] # 選取第一行
df =pd.DataFrame(np.random.rand(5,5),columns=list('ABCDE'))
df.iloc[0,:]
df.iloc[0,0] # 選取第一行的第一個(gè)元素
df = pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.iloc[0,0]
數(shù)據(jù)的清洗
df.columns = ['a','b'] # 重命名數(shù)據(jù)框的列名稱
df =pd.DataFrame({'A':np.array([1,np.nan,2,3,6,np.nan]),
????????????????????????????????'B':np.array([np.nan,4,np.nan,5,9,np.nan]),
????????????????????????????????'C':'foo'})
df.columns = ['a','b','c']
df
pd.isnull() # 檢查數(shù)據(jù)中空值出現(xiàn)的情況，并返回一個(gè)由布爾值(True,Fale)組成的列
df =pd.DataFrame({'A':np.array([1,np.nan,2,3,6,np.nan]),
????????????????????????????????'B':np.array([np.nan,4,np.nan,5,9,np.nan]),
????????????????????????????????'C':'foo'})
pd.isnull(df)
pd.notnull() # 檢查數(shù)據(jù)中非空值出現(xiàn)的情況，并返回一個(gè)由布爾值(True,False)組成的列
df =pd.DataFrame({'A':np.array([1,np.nan,2,3,6,np.nan]),
????????????????????????????????'B':np.array([np.nan,4,np.nan,5,9,np.nan]),
????????????????????????????????'C':'foo'})
pd.notnull(df)
df.dropna() # 移除數(shù)據(jù)框 DataFrame 中包含空值的行
df =pd.DataFrame({'A':np.array([1,np.nan,2,3,6,np.nan]),
????????????????????????????????'B':np.array([np.nan,4,np.nan,5,9,np.nan]),
????????????????????????????????'C':'foo'})
df.dropna()
df.dropna(axis=1) # 移除數(shù)據(jù)框 DataFrame 中包含空值的列
df =pd.DataFrame({'A':np.array([1,np.nan,2,3,6,np.nan]),
????????????????????????????????'B':np.array([np.nan,4,np.nan,5,9,np.nan]),
????????????????????????????????'C':'foo'})
df.dropna(axis=1)
df.dropna(axis=1,thresh=n) # 移除數(shù)據(jù)框df中空值個(gè)數(shù)不超過n的行
df =pd.DataFrame({'A':np.array([1,np.nan,2,3,6,np.nan]),
????????????????????????????????'B':np.array([np.nan,4,np.nan,5,9,np.nan]),
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 'C':'foo'})
test = df.dropna(axis=1,thresh=1)
test
df.fillna(x) # 將數(shù)據(jù)框 DataFrame 中的所有空值替換為 x
df =pd.DataFrame({'A':np.array([1,np.nan,2,3,6,np.nan]),
????????????????????????????????'B':np.array([np.nan,4,np.nan,5,9,np.nan]),
????????????????????????????????'C':'foo'})
df.fillna('Test')
s.fillna(s.mean()) -> 將所有空值替換為平均值
s = pd.Series([1,3,5,np.nan,7,9,9])
s.fillna(s.mean())
s.astype(float) # 將數(shù)組(Series)的格式轉(zhuǎn)化為浮點(diǎn)數(shù)
s = pd.Series([1,3,5,np.nan,7,9,9])
s.astype(float)
s.replace(1,'one') # 將數(shù)組(Series)中的所有1替換為'one'
s = pd.Series([1,3,5,np.nan,7,9,9])
s.replace(1,'one')
s.replace([1,3],['one','three']) # 將數(shù)組(Series)中所有的1替換為'one'，所有的3替換為'three'
s = pd.Series([1,3,5,np.nan,7,9,9])
s.replace([1,3],['one','three'])
df.rename(columns=lambda x: x + 2) # 將全體列重命名
df = pd.DataFrame(np.random.rand(4,4))
df.rename(columns=lambda x: x+ 2)
df.rename(columns={'old_name': 'new_name'}) # 將選擇的列重命名
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.rename(columns={'A':'newA','C':'newC'})
df.set_index('column_one') # 改變索引
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.set_index('B')
df.rename(index = lambda x: x+ 1) # 改變?nèi)w索引
df = pd.DataFrame(np.random.rand(10,5))
df.rename(index = lambda x: x+ 1)
數(shù)據(jù)的過濾(filter),排序(sort)和分組(groupby)
df[df[col] > 0.5] # 選取數(shù)據(jù)框df中對(duì)應(yīng)行的數(shù)值大于0.5的全部列
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df[df['A'] > 0.5]
df[(df[col] > 0.5) & (df[col] <0.7)] # 選取數(shù)據(jù)框df中對(duì)應(yīng)行的數(shù)值大于0.5，并且小于0.7的全部列
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df[(df['C'] > 0.5) & (df['D'] <0.7)]
df.sort_values(col1) # 按照數(shù)據(jù)框的列col1升序(ascending)的方式對(duì)數(shù)據(jù)框df做排序
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.sort_values('E')
df.sort_values(col2,ascending=False) # 按照數(shù)據(jù)框的列col2降序(descending)的方式對(duì)數(shù)據(jù)框df做排序
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.sort_values('A',ascending=False)
df.sort_values([col1,col2],ascending=[True,False]) # 按照數(shù)據(jù)框的列col1升序，col2降序的方式對(duì)數(shù)據(jù)框df做排序
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.sort_values(['A','E'],ascending=[True,False])
df.groupby(col) # 按照某列對(duì)數(shù)據(jù)框df做分組
df = pd.DataFrame({'A':np.array(['foo','foo','foo','foo','bar','bar']),
????????????????????????????????'B':np.array(['one','one','two','two','three','three']),
????????????????????????????????'C':np.array(['small','medium','large','large','small','small']),
????????????????????????????????'D':np.array([1,2,2,3,3,5])})
df.groupby('A').count()
df.groupby([col1,col2]) # 按照列col1和col2對(duì)數(shù)據(jù)框df做分組
df =pd.DataFrame({'A':np.array(['foo','foo','foo','foo','bar','bar']),
????????????????????????????????'B':np.array(['one','one','two','two','three','three']),
????????????????????????????????'C':np.array(['small','medium','large','large','small','small']),
????????????????????????????????'D':np.array([1,2,2,3,3,5])})
df.groupby(['B','C']).sum()
df.groupby(col1)[col2].mean() # 按照列col1對(duì)數(shù)據(jù)框df做分組處理后，返回對(duì)應(yīng)的col2的平均值
df =pd.DataFrame({'A':np.array(['foo','foo','foo','foo','bar','bar']),
????????????????????????????????'B':np.array(['one','one','two','two','three','three']),
????????????????????????????????'C':np.array(['small','medium','large','large','small','small']),
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 'D':np.array([1,2,2,3,3,5])})
df.groupby('B')['D'].mean()
df.pivot_table(index=col1,values=[col2,col3],aggfunc=mean)
# 做透視表，索引為col1,針對(duì)的數(shù)值列為col2和col3，分組函數(shù)為平均值
df =pd.DataFrame({'A':np.array(['foo','foo','foo','foo','bar','bar']),
????????????????????????????????'B':np.array(['one','one','two','two','three','three']),
????????????????????????????????'C':np.array(['small','medium','large','large','small','small']),
????????????????????????????????'D':np.array([1,2,2,3,3,5])})
df.pivot_table(df,index=['A','B'],columns=['C'],aggfunc=np.sum)
df.groupby(col1).agg(np.mean) #?返回按列col1分組的所有列的均值
df =pd.DataFrame({'A':np.array(['foo','foo','foo','foo','bar','bar']),
????????????????????????????????'B':np.array(['one','one','two','two','three','three']),
????????????????????????????????'C':np.array(['small','medium','large','large','small','small']),
????????????????????????????????'D':np.array([1,2,2,3,3,5])})
df.groupby('A').agg(np.mean)
df.apply(np.mean) # 對(duì)數(shù)據(jù)框df的每一列求平均值
df = pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.apply(np.mean)
df.apply(np.max,axis=1) # 對(duì)數(shù)據(jù)框df的每一行求最大值
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.apply(np.max,axis=1)
數(shù)據(jù)的連接(join)與組合(combine)
df1.append(df2) # 在數(shù)據(jù)框df2的末尾添加數(shù)據(jù)框df1，其中df1和df2的列數(shù)應(yīng)該相等
df1 = pd.DataFrame({'A': ['A0', 'A1', 'A2','A3'],
????????????????????????????????????'B': ['B0', 'B1', 'B2','B3'],
????????????????????????????????????'C': ['C0', 'C1', 'C2','C3'],
????????????????????????????????????'D': ['D0', 'D1', 'D2','D3']},index=[0, 1, 2, 3])
df2 = pd.DataFrame({'A': ['A4', 'A5', 'A6','A7'],
????????????????????????????????????'B': ['B4', 'B5', 'B6','B7'],
????????????????????????????????????'C': ['C4', 'C5', 'C6','C7'],
????????????????????????????????????'D': ['D4', 'D5', 'D6','D7']},index=[4, 5, 6, 7])
df1.append(df2)
pd.concat([df1, df2],axis=1) # 在數(shù)據(jù)框df1的列最后添加數(shù)據(jù)框df2,其中df1和df2的行數(shù)應(yīng)該相等
df1 = pd.DataFrame({'A': ['A0', 'A1', 'A2','A3'],
????????????????????????????????????'B': ['B0', 'B1', 'B2','B3'],
????????????????????????????????????'C': ['C0', 'C1', 'C2','C3'],
????????????????????????????????????'D': ['D0', 'D1', 'D2','D3']},index=[0, 1, 2, 3])
df2 = pd.DataFrame({'A': ['A4', 'A5', 'A6','A7'],
????????????????????????????????????'B': ['B4', 'B5', 'B6','B7'],
????????????????????????????????????'C': ['C4', 'C5', 'C6','C7'],
????????????????????????????????????'D': ['D4', 'D5', 'D6','D7']},index=[4, 5, 6, 7])
pd.concat([df1,df2],axis=1)
df1.join(df2,on=col1,how='inner') # 對(duì)數(shù)據(jù)框df1和df2做內(nèi)連接，其中連接的列為col1
df1 = pd.DataFrame({'A': ['A0', 'A1', 'A2','A3'],
???????????????????????????????????'B': ['B0', 'B1', 'B2','B3'],
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? 'key': ['K0', 'K1', 'K0','K1']})
df2 = pd.DataFrame({'C': ['C0', 'C1'],
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?'D': ['D0', 'D1']},index=['K0', 'K1'])
df1.join(df2, on='key')
數(shù)據(jù)的統(tǒng)計(jì)
df.describe() # 得到數(shù)據(jù)框df每一列的描述性統(tǒng)計(jì)
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.describe()
df.mean() # 得到數(shù)據(jù)框df中每一列的平均值
df = pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.mean()
df.corr() # 得到數(shù)據(jù)框df中每一列與其他列的相關(guān)系數(shù)
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.corr()
df.count() # 得到數(shù)據(jù)框df中每一列的非空值個(gè)數(shù)
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.count()
df.max() # 得到數(shù)據(jù)框df中每一列的最大值
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.max()
df.min() # 得到數(shù)據(jù)框df中每一列的最小值
df = pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.min()
df.median() # 得到數(shù)據(jù)框df中每一列的中位數(shù)
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.median()
df.std() # 得到數(shù)據(jù)框df中每一列的標(biāo)準(zhǔn)差
df =pd.DataFrame(np.random.rand(10,5),columns=list('ABCDE'))
df.std()
