DV 4주차(2)

SEABORN
Author

김보람

Published

September 28, 2022

import seaborn as sns
import matplotlib.pyplot as plt 
import numpy as np
import pandas as pd

Seaborn 특징

- 특1: 입력으로 데이터프레임을 선호한다. (matplotlib은 array를 선호)

  • 그렇다고 해서 데이터프레임이 아닌 경우 그림이 아예 안 그려지지는 않는다.
  • 데이터프레임 형태는 long form 과 wide form 이 있다. (ref:https://seaborn.pydata.org/tutorial/data_structure.html#long-form-vs-wide-form-data)
  • 참고로 long form이 더 우수한 저장형태!
  • wide-df = [array1,array2,array3]
  • long-df = [array_val, array_cat]

- 특2: matplotlib을 존경함. (ref:https://seaborn.pydata.org/)

sns boxplot

https://seaborn.pydata.org/generated/seaborn.boxplot.html

y1=[75,75,76,76,77,77,79,79,79,98] # A선생님에게 통계학을 배운 학생의 점수들
y2=[76,76,77,77,78,78,80,80,80,81] # B선생님에게 통계학을 배운 학생의 점수들

plt복습

plt.boxplot([y1,y2])
{'whiskers': [<matplotlib.lines.Line2D at 0x7f2a64e7a950>,
  <matplotlib.lines.Line2D at 0x7f2a64e7ac90>,
  <matplotlib.lines.Line2D at 0x7f2a64e910d0>,
  <matplotlib.lines.Line2D at 0x7f2a64e913d0>],
 'caps': [<matplotlib.lines.Line2D at 0x7f2a64e83050>,
  <matplotlib.lines.Line2D at 0x7f2a64e83350>,
  <matplotlib.lines.Line2D at 0x7f2a64e91710>,
  <matplotlib.lines.Line2D at 0x7f2a64e91a50>],
 'boxes': [<matplotlib.lines.Line2D at 0x7f2a64e7a650>,
  <matplotlib.lines.Line2D at 0x7f2a64e83d50>],
 'medians': [<matplotlib.lines.Line2D at 0x7f2a64e836d0>,
  <matplotlib.lines.Line2D at 0x7f2a64e91d90>],
 'fliers': [<matplotlib.lines.Line2D at 0x7f2a64e83a10>,
  <matplotlib.lines.Line2D at 0x7f2a64e9e110>],
 'means': []}

sns: wide df

df1=pd.DataFrame({1:y1,2:y2})
df1
1 2
0 75 76
1 75 76
2 76 77
3 76 77
4 77 78
5 77 78
6 79 80
7 79 80
8 79 80
9 98 81

- 예시1

sns.boxplot(data=df1)
<AxesSubplot:>

sns.boxplot(np.stack([y1,y2],axis=1))   # 잘 쓰진 않는데 된다
<AxesSubplot:>

sns: long df

value 를 넣고 그 value가 어떤 category에 있는지 넣는 방법

df2 = pd.DataFrame({'score': y1+y2, 'class': ['A']*len(y1)+['B']*len(y2)})
df2
score class
0 75 A
1 75 A
2 76 A
3 76 A
4 77 A
5 77 A
6 79 A
7 79 A
8 79 A
9 98 A
10 76 B
11 76 B
12 77 B
13 77 B
14 78 B
15 78 B
16 80 B
17 80 B
18 80 B
19 81 B

- 예시1

sns.boxplot(data=df2, x='class', y='score')
<AxesSubplot:xlabel='class', ylabel='score'>

sns: array

- 예시1

sns.boxplot(data=y1)
<AxesSubplot:>

- 예시2

sns.boxplot(y=y1)
<AxesSubplot:>

- 예시3

sns.boxplot(x=y1)
<AxesSubplot:>

sns histplot

- 데이터

x = np.random.randn(10000)
y = np.random.randn(10000) + 1

plt 복습

plt.hist(x,alpha=0.5)
plt.hist(y,alpha=0.5)
(array([2.000e+00, 1.500e+01, 1.550e+02, 7.670e+02, 2.062e+03, 3.085e+03,
        2.479e+03, 1.117e+03, 2.790e+02, 3.900e+01]),
 array([-3.5473064 , -2.74724651, -1.94718662, -1.14712673, -0.34706684,
         0.45299304,  1.25305293,  2.05311282,  2.85317271,  3.6532326 ,
         4.45329248]),
 <BarContainer object of 10 artists>)

- 예시2

plt.hist([x,y]);

sns: wide df

df1=pd.DataFrame({'x':x, 'y':y})
df1
x y
0 0.392340 -0.520932
1 -0.027382 2.332888
2 -0.266977 0.973511
3 -0.493336 2.801266
4 0.282255 0.433189
... ... ...
9995 -0.752878 2.394238
9996 -0.212005 2.293700
9997 -1.118235 2.660186
9998 1.558492 0.886679
9999 -0.753399 1.977537

10000 rows × 2 columns

- 예시

sns.histplot(data=df1);

sns.histplot(data=df1,bins=20)  # 칸 조정
<AxesSubplot:ylabel='Count'>

sns.histplot(data=df1,bins=20,kde=True);    # kde : 곡선

sns.histplot(data=df1,bins=20,kde=True,element="step");

sns.histplot(data=df1,bins=20,kde=True,element="step",lw=5) # mpl에 대한 존경심 확인 
<AxesSubplot:ylabel='Count'>

sns: long df

df2=pd.DataFrame({'val' : np.concatenate([x,y]), 'var': ['x']*len(x) + ['y']*len(y)})
df2
val var
0 0.392340 x
1 -0.027382 x
2 -0.266977 x
3 -0.493336 x
4 0.282255 x
... ... ...
19995 2.394238 y
19996 2.293700 y
19997 2.660186 y
19998 0.886679 y
19999 1.977537 y

20000 rows × 2 columns

sns.histplot(data=df2, x='val', hue='var', bins=20, kde=True, lw=0)
# hue:색깔 var로 구분하겠다! 
<AxesSubplot:xlabel='val', ylabel='Count'>

sns: array

sns.histplot(data=x)
<AxesSubplot:ylabel='Count'>

sns.histplot(x=x)
<AxesSubplot:ylabel='Count'>

sns.histplot(x=x, color='C0', bins=20, lw=0)
sns.histplot(x=y, color='C0', bins=20, lw=0)
<AxesSubplot:ylabel='Count'>

sns lineplot

np.random.seed(43052)
ϵ = np.random.randn(100)
y = np.cumsum(ϵ)  # 누적값 더하기

plt복습

plt.plot(ϵ,'--o')

plt.plot(y,'--o')

sns: array

sns.lineplot(data=ϵ)
<AxesSubplot:>

sns.lineplot(data=y)
<AxesSubplot:>

sns: wide df

df4 = pd.DataFrame({'ϵ' : ϵ , 'y' : y})
df4
ϵ y
0 0.383420 0.383420
1 1.084175 1.467595
2 1.142778 2.610373
3 0.307894 2.918267
4 0.237787 3.156054
... ... ...
95 1.308688 -10.598788
96 0.405376 -10.193412
97 -0.185070 -10.378481
98 1.055388 -9.323094
99 1.187014 -8.136079

100 rows × 2 columns

sns.lineplot(data=df4)
<AxesSubplot:>

sns.lineplot(data=df4, dashes=False)   # dashes : 둘다 실선
<AxesSubplot:>

sns: long df

df5= pd.DataFrame({'idx':list(range(100))*2,'val':np.concatenate([ϵ,y]),'cat':['eps']*100 + ['y']*100 })
df5
idx val cat
0 0 0.383420 eps
1 1 1.084175 eps
2 2 1.142778 eps
3 3 0.307894 eps
4 4 0.237787 eps
... ... ... ...
195 95 -10.598788 y
196 96 -10.193412 y
197 97 -10.378481 y
198 98 -9.323094 y
199 99 -8.136079 y

200 rows × 3 columns

sns.lineplot(data=df5, x='idx',y='val',hue='cat')
<AxesSubplot:xlabel='idx', ylabel='val'>

sns.lineplot(data=df5, x='idx',y='val',style='cat',hue='cat',markers=True)
<AxesSubplot:xlabel='idx', ylabel='val'>

sns.lineplot(data=df5, x='idx',y='val',style='cat',hue='cat',dashes=[(3,1),(3,3)],markers=['o','o'])
<AxesSubplot:xlabel='idx', ylabel='val'>

숙제

y1 = np.random.randn(90).cumsum() 
y2 = np.random.randn(120).cumsum()
plt.plot(y1,'--o')
plt.plot(y2,'--o')

y = ([y1,y2])
sns.lineplot(data=y, markers=['o','o'])
<AxesSubplot:>