My pandas (python)

Summary for myself Updated from time to time Describe the command you used and examined ** I just need to know myself, so the term may be wrong in some places **

!! View

#command
    #Argument option description

Module loading

import pandas 

Pandas data objects and frequent operations

Series

#One-dimensional data object
#Array in my image
ser = pandas.Series()

DataFrame

#Two-dimensional data object
#My image is similar to a DB table
df = pandas.DataFrame()

Manipulating the DataFrame structure

#x,y,...Sort in order of
df.sort_values(x,y,...)

#Removed the argument index.
df.drop(x)
    axis = 1 #Delete column

#Combine data frames
df.merge(x,y,on=z) #join x and y tables with z column as key
    suffixes=() #Suffix to be added when there are duplicate items Separated by commas, the first is the left df suffix, and the second is the right suffix.
#Swap rows and columns
df.transpose()

df.concat([x,y,z,...])
    #A list of dfs you want to combine into a list of arguments

DataFrame extraction / acquisition operation

#Maximum value
df.max()

#minimum value
df.min()

#Extract information for each item in the data frame
df.info()
    #No special arguments are required

#Extract by row number / column number
df.iloc[line,Column] #Argument:And all

#Extract by row name / column name
df.loc[line,Column] #Argument:And all

#WHERE IN in SQL
df.isin()
    #Arguments are lists, etc.

#Returns the record for the argument from the beginning
df.head()

#Get median value
df.median()

#Replace Nan
df.filna()

#Get summary statistics
df.describe() 
#Return the following statistics in DataFrame for all numeric columns
 #count:Number of elements
 #unique:Number of unique (unique) value elements
 #top:Mode
 #freq:Mode frequency (number of occurrences)
 #mean:Arithmetic mean
 #std:standard deviation
 #min:minimum value
 #max:Maximum value
 #50%:Median
 #25%: 1/Quartile
 #75%: 3/Quartile

Groupby

#groupby
group = df.groupby()
    as_index=False #If False, the reference value of aggregation will not be an index
    how = left,right,outer
    #Item name in argument

#number
group.size()

#Aggregate specific items in various ways
df.agg({'Items to be aggregated':['Aggregation method list']

DataFrame read operation

#Read csv. This is when reading data whose delimiter is a comma
df.read_csv()
　　encoding: #Specify the character code
　　header= #Set what row the column name is
    name= #Set column name
    dtype= #Specify data type with dictionary type
    sep= #Specifying the delimiter
    engine=
    usecols = #Specify the column to read in the list.

#Read table. This is when the delimiter reads the tab data
df.read_table()
　　encoding: #Specify the character code
　　header= #Set what row the column name is
    name= #Set column name

#Have the DB read
df.read_sql()
    #The first argument is SQL
    #The second argument is the connection object

DataFrame export operation

df.to_csv()
    encoding= #Character code
    index= #Output index together or default is True

DataFrame row name / column name operation

#Overwrite column name
df.columns = [list]
df.rename(columns={Current column name:New column name})

#Overwrite index
df.index = [list]

#Column name / index name change
df.rename({Current name: New name})
    axis=1 #Change column name. If not specified, it will be the line name.
    

#Reindex
df.reset_index()
    drop=True #Delete existing index

DataFrame write operation

#Insert line
df.[Column name] = x


#Replace
df.replace({Current character:New character}) #The argument is dictionary type{Character to replace:Character after replacement}

#Insert argument as new line, insert line is list, Series, numpy.array
df.append() 

#Add column
df.assign()

#Apply function to each column / row
df.apply()
    axis=1 #Line by line
    axis=0 #By column
    #Function as an argument. Lambda style is fine.

Turn for statement in DataFrame

#Extract the DataFrame line by line and apply it to for.
for index,row in df.iterrows()
  #The return value is index and other row elements

Lambda expression in DataFrame

This article is very easy to understand. ↓ Recursion Substitution Eradication Committee for Data Processing in Python / pandas

Graph from pandas

#bar graph
df.plot.bar()

Do something about Nan

#Distinguishing Nan
df.isnull()


#Remove Nan
df.dropna()
    axis=1 #Delete column.

#Replace nan
df.fillna()
 

Manage the duplication.

#Find duplicate lines
#The return value is True if it overlaps with the index, otherwise False column
df.dupulicated()
    keep = False #If you do not specify, you cannot retrieve it as a duplicate.
#Remove completely duplicate lines
df.drop_duplicates()

drawing

#Pair plot
grr = pd.scatter_matrix(df)
    #df is the data to plot
    c= #Value to scale
    figsize=(x,y) #Figure size
    marker= #Marker shape
    hist_kwds={} #Histogram settings
    s= #Marker size
    alpha= #Transparency