DAY45. Python Matplot (3)Seaborn

Seaborn 패키지

다양한 배경 테마 제공
통계용 차트 제공
자체 dataset 제공

import seaborn as sn #별칭

제공 dataset 확인

print(sn.get_dataset_names())

['anagrams', 'anscombe', 'attention', 'brain_networks', 'car_crashes', 'diamonds', 'dots', 'exercise', 'flights', 'fmri', 'gammas', 'geyser', 'iris', 'mpg', 'penguins', 'planets', 'taxis', 'tips', 'titanic']

len(sn.get_dataset_names()) # 19

2. dataset load

iris = sn.load_dataset(name='iris')
type(iris) #pandas.core.frame.DataFrame
iris.info()
iris.head()

tips = sn.load_dataset(name='tips')
tips.info()

titanic = sn.load_dataset(name='titanic')
titanic.info()

flights = sn.load_dataset(name='flights')
flights.info()

 

 

 

 

 

object vs category

object : 순서 변경이 불가능한 문자열
category : 순서 변경이 가능한 문자열

import seaborn as sn

1. object vs category
dataset load

tips = sn.load_dataset(name = 'tips')
titanic = sn.load_dataset(name = 'titanic')

titanic.info()

subset 생성

df = titanic[['survived', 'age', 'class', 'who']]
df.info()

 2   class     891 non-null    category
 3   who       891 non-null    object
  
category 정렬

 df.head()

 0   survived  891 non-null    int64   
 1   age       714 non-null    float64 
 2   class     891 non-null    category
 3   who       891 non-null    object
 
object 정렬

df.sort_values(by = 'who') #오름차순
df['who'].unique() #array(['man', 'woman', 'child'], dtype=object)

'child' > 'man' > 'woman' : child(base)
object -> category
'man' > 'woman' > 'child' : man(base)

category 수정 : DF['column'].astype('type')

df['who_new'] = df['who'].astype('category')
df.head()
df.info()

 3   who       891 non-null    object  
 4   who_new   891 non-null    category
  
범주 순서 변경

df['who_new'] = df['who_new'].cat.set_categories(['man', 'woman', 'child'])

R에서는

df$who_new = fator(df$who_new, levels = c('man', 'woman', 'child'))

category 정렬

df.sort_values(by = 'who_new')

2. 범주형 자료 시각화
1) 그래프 배경 스타일

sn.set_style(style = 'darkgrid')

2) 범주형 자료 시각화

tips.head()
sn.countplot(x = 'smoker', data = tips) #2개 범주 
sn.countplot(x = 'class', data = titanic) #3개 범주
sn.countplot(x = 'day', data = tips) #4개 범주

 

 

 

 

 

continous

연속형 변수 시각화 
히스토그램, 산점도, 산점도 행렬, boxplot

import matplotlib.pyplot as plt
import seaborn as sn

dataset load 

iris = sn.load_dataset('iris')
tips = sn.load_dataset('tips')
type(iris) # pandas.core.frame.DataFrame

 

1. 히스토그램

iris.info()
x = iris.sepal_length # iris['sepal_length']

sn.histplot(data=x)
plt.title('sepal_length hist')
plt.show()

 

2. distplot : hist + kde 

sn.distplot(x, bins=20, hist=True, kde=True)
plt.title('sepal_length hist & kde')
plt.show()

kde : 커널밀도추정 

 

3. 산점도 행렬 
sn.pairplot(data = DataFrame, hue='집단변수') 

sn.pairplot(data = iris, hue='species') 
plt.show()

 

 

4. 산점도 : 연속형 vs 연속형 

sn.scatterplot(x='sepal_length', y='petal_length',
               data = iris)
plt.show()

 

hue='집단변수'

sn.scatterplot(x='sepal_length', y='petal_length',
               hue='species', data = iris)
plt.show()

5. boxplot

tips.info()
sn.boxplot(x='day', y='total_bill',
           hue='sex', data=tips)
plt.show()

 

 

 

 

 

통계적 모델 관련 시각화 

시계열분석, 상관분석, 회귀분석, 분류분석  

import matplotlib.pyplot as plt
import seaborn as sn 
import pandas as pd  # object

 

dataset load 

iris = sn.load_dataset('iris')
tips = sn.load_dataset('tips')
flights = sn.load_dataset('flights')

iris.info()
tips.info()
flights.info()

 

1. 오차대역폭을 갖는 시계열 시각화 
x : 시간축, y : 통계량 

sn.lineplot(x='year', y='passengers',
            data = flights) 
plt.show()

hue 추가 : 월 단위 탑승객 추세 

sn.lineplot(x='year', y='passengers',
            hue='month', data = flights) 
plt.show()
print(flights)

2. 상관분석 : 연속형 vs 연속형 vs 이산형   

sn.relplot(x='total_bill', y='tip', 
           size='size', data = tips)
plt.show()

상관계수 : 0.6757341092113641

tips['total_bill'].corr(tips['tip'])

 

3. 선형회귀분석 : 산점도와 회귀선, 신뢰구간(ci=95)  

sn.lmplot(x='sepal_length', y='petal_length', 
          data=iris, ci=95)
plt.show()

hue 속성 : species

sn.lmplot(x='sepal_length', y='petal_length', 
          data=iris, hue='species',  ci=95)
plt.show()

4. 분류분석 : 혼동행렬(confusion matrix)
1) y true vs y pred 

y_true = pd.Series([1,0,1,0,0])
y_pred = pd.Series([1,0,1,0,1])

2) 혼동행렬 : 교차분할표 

mat = pd.crosstab(index=y_true, columns=y_pred)
mat

col_0  0  1
row_0      
0      2  1
1      0  2

type(mat) # pandas.core.frame.DataFrame

 

3) 분류정확도 

score = (mat.iloc[0,0] + mat.iloc[1,1]) / len(y_true) 
print('분류정확도 = ', score) # 분류정확도 =  0.8

4) heatmap 

sn.heatmap(data=mat, annot=True, fmt=".2f", linewidths=2)

annot=True : label 표시 

plt.ylabel("Real label")
plt.xlabel("Predicted label")
acc_score = f"Accuracy score : {score}"
plt.title(acc_score, size = 15)
plt.show()