Pandas 데이터분석 기초 실습-7

팬더스 apply 함수 다양한 활용 방법¶

In [1]:

import pandas as pd
date_list = [{'yyyy-mm-dd':'2000-06-27'},
{'yyyy-mm-dd':'2002-09-24'},
{'yyyy-mm-dd':'2005-12-20'}]
df = pd.DataFrame(date_list,columns=['yyyy-mm-dd'])
df

Out[1]:

	yyyy-mm-dd
0	2000-06-27
1	2002-09-24
2	2005-12-20

In [2]:

#column 추가하기 (year)
def extract_year(column):
    return column.split('-')[0]

In [3]:

df['year'] = df['yyyy-mm-dd'].apply(extract_year)
df

Out[3]:

	yyyy-mm-dd	year
0	2000-06-27	2000
1	2002-09-24	2002
2	2005-12-20	2005

In [4]:

#parameter 넣기
def get_age(year,current_year):
    return current_year - int(year)

In [5]:

df['age'] = df['year'].apply(get_age,current_year = 2018)
df
#키워드 인수로 넣기

Out[5]:

	yyyy-mm-dd	year	age
0	2000-06-27	2000	18
1	2002-09-24	2002	16
2	2005-12-20	2005	13

In [6]:

def get_introduce(age,prefix,suffix):
    return prefix + str(age) + suffix

In [7]:

df['introduce'] = df['age'].apply(get_introduce,prefix ='I am ',suffix = ' years old.')
df

Out[7]:

	yyyy-mm-dd	year	age	introduce
0	2000-06-27	2000	18	I am 18 years old.
1	2002-09-24	2002	16	I am 16 years old.
2	2005-12-20	2005	13	I am 13 years old.

In [8]:

#여러개의 column값을 apply function사용
def get_introduce_2(row):
    return 'I was born in '+ str(row.year) + ' my age is '+ str(row.age)

In [9]:

df.introduce = df.apply(get_introduce_2, axis = 1)  # == df['introduce']
# axis = 1을 활용함으로 써 row 에 있는 모든 데이터 활용
df

Out[9]:

	yyyy-mm-dd	year	age	introduce
0	2000-06-27	2000	18	I was born in 2000 my age is 18
1	2002-09-24	2002	16	I was born in 2002 my age is 16
2	2005-12-20	2005	13	I was born in 2005 my age is 13

map, applymap 함수활용¶

In [10]:

import pandas as pd

date_list = [{'date':'2000-06-27'},
{'date':'2002-09-24'},
{'date':'2005-12-20'}]

df = pd.DataFrame(date_list,columns=['date'])
df

Out[10]:

	date
0	2000-06-27
1	2002-09-24
2	2005-12-20

In [11]:

def extract_year(date):
    return date.split('-')[0]

In [12]:

df['year'] = df['date'].map(extract_year)
# 연도 만 끄집어 내서 date column 만들기
df

Out[12]:

	date	year
0	2000-06-27	2000
1	2002-09-24	2002
2	2005-12-20	2005

In [13]:

job_list = [{'age':20,'job':'student'},
{'age':30,'job':'developer'},
{'age':30,'job':'teacher'}]
df = pd.DataFrame(job_list)
df

Out[13]:

	age	job
0	20	student
1	30	developer
2	30	teacher

In [14]:

#머신러닝의 경우 string을 숫자로 바꿔 줘야할 경우가 생긴다
df.job = df.job.map({'student':1,'developer':2,'teacher':3})
#딕셔너리 전달을 통한 변경
df

Out[14]:

	age	job
0	20	1
1	30	2
2	30	3

applymap 함수¶

하나의 column 변경이 아닌 모든 dataframe의 변경을 원하면¶

applymap 함수를 사용한다¶

In [15]:

x_y = [{'x':5.5,'y':-5.6,'z':1.1},
{'x':-5.2,'y':5.5,'z':-2.2},
{'x':-1.6,'y':-4.5,'z':-3.3}]
df =pd.DataFrame(x_y)
df

Out[15]:

	x	y	z
0	5.5	-5.6	1.1
1	-5.2	5.5	-2.2
2	-1.6	-4.5	-3.3

In [16]:

import numpy as np

In [17]:

df = df.applymap(np.around) #반올림, 값의 일괄적인 적용
df

Out[17]:

	x	y	z
0	6.0	-6.0	1.0
1	-5.0	6.0	-2.0
2	-2.0	-4.0	-3.0

column내 unique 한 값 뽑아내고 개수 확인하기¶

컬럼에 여러 값이 있을 때, 중복 없이 어떤 값들이 있는 지 확인하는 방법입니다.¶

In [18]:

import pandas as pd
job_list = [{'name':'John','job':'teacher'},
{'name':'Nate','job':'teacher'},
{'name':'Fred','job':'teacher'},
{'name':'Abraham','job':'student'},
{'name':'Brian','job':'student'},
{'name':'Janny','job':'developer'},
{'name':'Nate','job':'teacher'},
{'name':'Ian','job':'teacher'},
{'name':'Chris','job':'banker'},
{'name':'Phillip','job':'lawyer'},
{'name':'Phillip','job':'basketball player'},
{'name':'Gwen','job':'teacher'},
{'name':'Jessy','job':'student'}]

df = pd.DataFrame(job_list,columns =['name','job'])
df

Out[18]:

	name	job
0	John	teacher
1	Nate	teacher
2	Fred	teacher
3	Abraham	student
4	Brian	student
5	Janny	developer
6	Nate	teacher
7	Ian	teacher
8	Chris	banker
9	Phillip	lawyer
10	Phillip	basketball player
11	Gwen	teacher
12	Jessy	student

In [19]:

#unique 한 값 뽑기
df.job.unique()

Out[19]:

array(['teacher', 'student', 'developer', 'banker', 'lawyer',
       'basketball player'], dtype=object)

In [20]:

df.job.value_counts()   #각각의 unique한 value에 몇개가 있는지 추출

Out[20]:

teacher              6
student              3
developer            1
banker               1
lawyer               1
basketball player    1
Name: job, dtype: int64

두개의 데이터프레임 합치기¶

row로 합치기¶

In [21]:

l1 = [{'name': 'John', 'job': "teacher"},
      {'name': 'Nate', 'job': "student"},
      {'name': 'Fred', 'job': "developer"}]

l2 = [{'name': 'Ed', 'job': "dentist"},
      {'name': 'Jack', 'job': "farmer"},
      {'name': 'Ted', 'job': "designer"}]
df1 = pd.DataFrame(l1, columns = ['name', 'job'])
df2 = pd.DataFrame(l2, columns = ['name', 'job'])

In [22]:

df1

Out[22]:

	name	job
0	John	teacher
1	Nate	student
2	Fred	developer

In [23]:

df2

Out[23]:

	name	job
0	Ed	dentist
1	Jack	farmer
2	Ted	designer

In [24]:

#첫번째 방법
result  = pd.concat([df1,df2],ignore_index=True) #2개의 데이터프레임을 list 형식으로 넣기
result

Out[24]:

	name	job
0	John	teacher
1	Nate	student
2	Fred	developer
3	Ed	dentist
4	Jack	farmer
5	Ted	designer

In [25]:

#두번째 방법
result = df1.append(df2,ignore_index =True)
result

Out[25]:

	name	job
0	John	teacher
1	Nate	student
2	Fred	developer
3	Ed	dentist
4	Jack	farmer
5	Ted	designer

두번째 데이터프레임을 첫번째 데이터프레임의 새로운 컬럼(열)으로 합치기¶

In [26]:

l3 = [{'name': 'John', 'job': "teacher"},
      {'name': 'Nate', 'job': "student"},
      {'name': 'Jack', 'job': "developer"}]

l4 = [{'age': 25, 'country': "U.S"},
      {'age': 30, 'country': "U.K"},
      {'age': 45, 'country': "Korea"}]
df1 = pd.DataFrame(l3,columns= ['name','job'])
df2 = pd.DataFrame(l4,columns=['age','country'])

In [27]:

df1

Out[27]:

	name	job
0	John	teacher
1	Nate	student
2	Jack	developer

In [28]:

df2

Out[28]:

	age	country
0	25	U.S
1	30	U.K
2	45	Korea

In [33]:

result = pd.concat([df1,df2],axis = 1,ignore_index= True)
result.columns = ['name','job','age','country']
result    

Out[33]:

	name	job	age	country
0	John	teacher	25	U.S
1	Nate	student	30	U.K
2	Jack	developer	45	Korea

두개의 리스트 합치기¶

machine learning에서 실제값과 예측값 비교할때 많이 쓰임¶

제대로 예측이 됐는지?¶

In [34]:

label = [1,2,3,4,5]
prediction=[1,2,2,4,4]

In [36]:

comparison =pd.DataFrame({'label':label,'prediction' : prediction})
comparison

Out[36]:

	label	prediction
0	1	1
1	2	2
2	3	2
3	4	4
4	5	4