PYTHON PANDAS Cheat Sheet

by sanjeev95 via cheatography.com/111141/cs/21621/

install and import Rows and columns Data Cleaning

installing pandas to delete a row - [axis=0 means rows] df.set​_in​dex​('c​olu​mn_​‐ Change the
pip install pandas new_df = df.dro​p([​2,3​],axis = 0) one') index with a new

pip install pandas #this drops the row with index column

import pandas as pd 2,3 df.columns = ['new_​col​‐ Rename

to delete a column- [axis=1 means columns] _na​me1​','​new​_co​l_n​‐ columns
Reading and describing new_df = df.dro​p([​'co​l1'​,'c​‐ ame​2',​'ne​w_c​ol_​‐
ol2​'],axis = 0) #this drops the name3']
pd -> pandas
df-> dataframe column with name col1 and col2 pd.isn​ull() Checks for null
Values, Returns
to read a file into a dataframe
Df manipu​lation Boolean Arrray
df= pd.rea​d_c​sv(​'fi​len​ame')
create or edit a new column pd.not​null() Opposite of
look at the first 5 lines
df['n​ew_​col​name'] = 5 #this pd.isn​ull()
df.he​ad()
creates a new new column with df.dro​pna() Drop all rows
to describe df
all values as 5 that contain null
df.de​scr​ibe()
values
create a new column
df.in​fo()
df['n​ew_​col​name'] = [list of df.dro​pna​(ax​is=1) Drop all
to print all the column names columns that
values] #this creates a new
telecom_data.columns
column with list of values contain null
to get the dimension of df values
assigned to each corres​ponding
df.shape row df.dro​pna​(ax​is=​1,t​hre​‐ Drop all rows
sh=n) have have less
NOTE : df['n​ew_​col​name'] = [list
Sorting and filtering than n non null
of values] throws an error if the no of
sort df.fil​lna(x) Replace all null
items in [list of values] doesn't match no of
sorting can be done column wise - default is values with x
rows
ascending
create or edit a new row
df.so​rt_​val​ues​(by​='Total day JOIN/C​OMBINE
df.lo​c[i​nde​x_o​f_row] = [list of
charge') df1.a​ppe​nd(​df2) Adds the rows in df1
items]
df.so​rt_​val​‐ Sort values by col1 in to the end of
NOTE : df.lo​c[i​nde​x_o​f_row] = df2 (columns should
ues​(col1) ascending order (use
ascending =False for [list of items] throws an error if the be identical)
descending sort) no of items doesn't match no of rows
pd.co​nca​t([df1, Adds the columns in
df.so​rt_​val​‐ Sort values by col1 in df2],a​xis=1) df1 to the end
Selection
ues​([c​ol1​,co​‐ ascending order then of df2 (rows should
l2]​,as​cen​din​‐ col2 in descending be

g=[​Tru​e,F​‐ order identical)

alse]) df1.j​oin​(df​2,o​‐ joins the columns in

n=c​ol1​,ho​w='​‐ df1 with the columns
Filtering
inn​er') on
df[co​ndi​tion] ​ ​ #eg:
df2 where the rows
df[df[​'co​l']​>5]
for col have identical
df[df​['col'] Rows where the values. how can be
> 0.5] column col is greater one of 'left',
than 0.5 'right', 'outer', 'inner'
df[(d​f[col] > Rows where 0.7 > col left = takes the index of left df
0.5) & > 0.5 right =takes the index of left
(df[col] < outer = union of both keys
0.7)] inner = inters​ection of both keys

Inplace
NOTE df[col] Returns column with label
df.me​rge​(df2) gives you a copy of df col as Series

merged with df2. you may save it to a new df[[col1, Returns multiple columns as
variable. ex df3=d​f.m​erg​e(df2) col2]] a new DataFrame

if you want to merge df2 to df right away use ​ Country ​ ​Capital ​ ​ ​ ​ ​ ​Pop​ulation
inplace. df.me​rge​(df​2,i​npl​ace​=True) 1 Belgiu​m Brussels ​ ​ ​ ​111​90846
2 India ​ ​ ​ ​ ​ ​ New Delhi ​130​3171035
3 Brazil ​ ​ ​ ​ ​ ​Bra​silia ​ ​ ​ ​ ​ ​207​847528
df.il​oc([0], [0]) --> 'Belgium' |
s.iloc[0] | Selection by position (0th position
on row and column)
df.lo​c([0], ['Coun​try']) -->
'Belgium'
df.ix[2] -->
Country ​ ​ ​ ​ ​ Brazil
Capital Brasilia
Population 207847528
df.ix[1, 'Capit​al'] --> 'New Delhi'

df.ilo​c[0,:] | select First row

By sanjeev95 Published 20th January, 2020. Sponsored by Readable.com

cheatography.com/sanjeev95/ Last updated 22nd January, 2020. Measure your website readability!
Page 1 of 2. https://readable.com