[Kaggle] HeartAttack 예측

220131 작성

<본 블로그는 Kaggle 을 참고해서 공부하며 작성하였습니다>

https://www.kaggle.com/fahadmehfoooz/heartattack-prediction-with-91-8-accuracy/notebook

HeartAttack prediction with 91.8 % Accuracy

 

 

 

 

 

 

1. 데이터 로드

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

from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler

from sklearn.tree import DecisionTreeClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import classification_report
from sklearn import svm
from sklearn.svm import SVC
from sklearn.linear_model import SGDClassifier
from sklearn.naive_bayes import BernoulliNB, GaussianNB
from xgboost import XGBClassifier # model

from sklearn.ensemble import VotingClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import AdaBoostClassifier
from sklearn.ensemble import RandomForestRegressor

from sklearn.metrics import confusion_matrix, classification_report
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, roc_auc_score
from sklearn.metrics import roc_curve, auc

heart = pd.read_csv("heart.csv")
heart.head()

heart.info()

age : Age of the patient
sex : Sex of the patient
cp : Chest pain type
- 0 = Typical Angina
- 1 = Atypical Angina
- 2 = Non-anginal Pain (비협심증) 
- 3 = Asymptomatic (무증상)

trtbps : Resting blood pressure (in mm Hg)
chol : Cholestoral in mg/dl fetched via BMI sensor
fbs : (fasting blood sugar > 120 mg/dl) 
- 1 = True 
- 0 = False

restecg : Resting electrocardiographic results
- 0 = Normal
- 1 = ST-T wave normality
- 2 = Left ventricular hypertrophy

thalachh : Maximum heart rate achieved
oldpeak : Previous peak
slp : Slope
caa : Number of major vessels
thall : Thalium Stress Test result ~ (0,3)
exng : Exercise induced angina ~ 
- 1 = Yes 
- 0 = No

- output : Target variable

 

 

• NULL 없음

heart.isnull().sum()

 

• 중복값 있나?

heart[heart.duplicated()]

 

 

• 하나 있으니 DROP!

heart.drop_duplicates(keep = "first", inplace = True)

 

 

•  요약!

heart.describe()

 

 

 

 

• 상관 관계

heart.corr()

 

 

 

 

 

 

2. 데이터 시각화

• 남녀 비율

sex = heart.sex.value_counts()

p = sns.countplot(data = heart, x = "sex")
plt.show()

 

 

• 질병 유형

cp = heart.cp.value_counts()

p = sns.countplot(data = heart, x = "cp")
plt.show()

 

 

 

• 공복 혈당 (공복 시 혈액 내 당 농도)

fbs = heart.fbs.value_counts()

p = sns.countplot(data = heart, x = "fbs")
plt.show()

 

 

 

 

• Resting electrocardiographic results

restecg = heart.restecg.value_counts()

p = sns.countplot(data = heart, x = "restecg")
plt.show()

 

 

 

• Exercise induced angina ~ 

exng = heart.exng.value_counts()

p = sns.countplot(data = heart, x = "exng")
plt.show()

 

 

• Thalium Stress Test result ~ (0,3)

thall = heart.thall.value_counts()

p = sns.countplot(data = heart, x = "thall")
plt.show()

 

 

 

• 나이

plt.figure(figsize=(10,10))

sns.displot(heart.age, color="red", label="Age", kde= True)
plt.legend()

 

 

 

• Resting Blood Pressure

plt.figure(figsize=(20,20))

sns.displot(heart.trtbps , color="green", label="Resting Blood Pressure", kde= True)
plt.legend()

 

 

 

• Attack VS Age

plt.figure(figsize=(10,10))
sns.distplot(heart[heart['output'] == 0]["age"], color='green',kde=True,) 
sns.distplot(heart[heart['output'] == 1]["age"], color='red',kde=True)
plt.title('Attack VS Age')
plt.show()

 

 

 

• Trtbs VS Age

plt.figure(figsize=(10,10))

sns.distplot(heart[heart['output'] == 0]["trtbps"], color='orange',kde=True,) 
sns.distplot(heart[heart['output'] == 1]["trtbps"], color='red',kde=True)
plt.title('Trtbs versus Age')
plt.show()

 

 

 

 

• Thalachh VS Age

plt.figure(figsize=(10,10))
sns.distplot(heart[heart['output'] == 0]["thalachh"], color='pink',kde=True,) 
sns.distplot(heart[heart['output'] == 1]["thalachh"], color='red',kde=True)
plt.title('Thalachh versus Age')
plt.show()

 

 

 

 

• Pair Plot

plt.figure(figsize=(20,20))
sns.pairplot(heart)
plt.show()

 

 

 

 

• Violin Plot

plt.figure(figsize=(13,13))

plt.subplot(2,3,1)
sns.violinplot(x = 'sex', y = 'output', data = heart)
plt.subplot(2,3,2)
sns.violinplot(x = 'thall', y = 'output', data = heart)
plt.subplot(2,3,3)
sns.violinplot(x = 'exng', y = 'output', data = heart)
plt.subplot(2,3,4)
sns.violinplot(x = 'restecg', y = 'output', data = heart)
plt.subplot(2,3,5)
sns.violinplot(x = 'cp', y = 'output', data = heart)
plt.xticks(fontsize=9, rotation=45)
plt.subplot(2,3,6)
sns.violinplot(x = 'fbs', y = 'output', data = heart)

plt.show()

 

 

 

 

 

3. 데이터 전처리

heart.head()

 

 

 

 

• output 따로

X = heart.drop(['output'], axis = 1) # 칼럼으로
y = heart["output"]

 

•  train, test 분리

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size= 0.2, random_state= 42)

print('Shape for training data', X_train.shape, y_train.shape)
print('Shape for testing data', X_test.shape, y_test.shape)

 

 

 

 

• scaling

scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)

 

 

 

 

 

4. 모델

•  Logistic Regression

logistic = LogisticRegression()
logistic.fit(X_train, y_train)

predicted = logistic.predict(X_test)
conf = confusion_matrix(y_test, predicted)

print ("logistic Confusion Matrix : \n", conf)
print ("The accuracy of Logistic Regression is : ", accuracy_score(y_test, predicted)*100, "%")

 

 

 

• Gaussian Naive Bayes
• 각 특성을 개별로 취급해 파라미터를 학습하고 그 특성에서 클래스별 통계를 단순하게 취합

gaussian = GaussianNB()
gaussian.fit(X_train, y_train)
  
predicted = gaussian.predict(X_test)
conf = confusion_matrix(y_test, predicted)

print ("GaussianNB Confusion Matrix : \n", conf)
print("The accuracy of Gaussian Naive Bayes model is : ", accuracy_score(y_test, predicted)*100, "%")

 

 

 

• Bernoulli Naive Bayes

bernoulli = BernoulliNB()
bernoulli.fit(X_train, y_train)
  
predicted = bernoulli.predict(X_test)
conf = confusion_matrix(y_test, predicted)

print ("Bernoulli Confusion Matrix : \n", conf)
print("The accuracy of Bernoulli Naive Bayes model is : ", accuracy_score(y_test, predicted)*100, "%")

 

 

 

• SVM

svm = SVC()
svm.fit(X_train, y_train)
  
predicted = svm.predict(X_test)
conf = confusion_matrix(y_test, predicted)

print ("SVM Confusion Matrix : \n", conf)
print("The accuracy of SVM is : ", accuracy_score(y_test, predicted)*100, "%")

 

 

 

 

• Random Forest

rf = RandomForestRegressor(n_estimators = 100, random_state = 42)  
rf.fit(X_train, y_train)
  
predicted = rf.predict(X_test)

print("The accuracy of Random Forest is : ", accuracy_score(y_test, predicted.round())*100, "%")

 

 

 

• K Nearest Neighbours

KNN = KNeighborsClassifier(n_neighbors = 1)  
KNN.fit(X_train, y_train)

predicted = KNN.predict(X_test)
conf = confusion_matrix(y_test, predicted)

print ("KNN Confusion Matrix : \n", conf)
print("The accuracy of KNN is : ", accuracy_score(y_test, predicted.round())*100, "%")

 

 

 

• +) KNN 최적화

error_rate = []
  
for i in range(1, 40):
      
    KNN = KNeighborsClassifier(n_neighbors = i)
    KNN.fit(X_train, y_train)
    pred_i = KNN.predict(X_test)
    error_rate.append(np.mean(pred_i != y_test))
  
plt.figure(figsize =(10, 6))
plt.plot(range(1, 40), error_rate, color ='blue',
                linestyle ='dashed', marker ='o',
         markerfacecolor ='red', markersize = 10)
  
plt.title('Error Rate vs. K Value')
plt.xlabel('K')
plt.ylabel('Error Rate')

KNN_after = KNeighborsClassifier(n_neighbors = 7)
KNN_after.fit(X_train, y_train)

predicted = KNN_after.predict(X_test)
conf = confusion_matrix(y_test, predicted)

print ("KNN Confusion Matrix : \n", conf)
print("The accuracy of KNN is : ", accuracy_score(y_test, predicted.round())*100, "%")

 

 

 

 

• K Gradient Boosting
• 여러 개의 결정 트리를 임의적으로 학습하는 앙상블의 부스팅 유형

model = XGBClassifier(use_label_encoder=False)
model.fit(X_train, y_train)
   
predicted = model.predict(X_test)
conf = confusion_matrix(y_test, predicted)

print ("XGBClassifier Confusion Matrix : \n", conf)
print ("The accuracy of XGBClassifier is : ", accuracy_score(y_test, predicted)*100, "%")

 

 

 

 

 

5. 결론

knn 이 높당

참고한 kaggle 에선 svm 이 높았눈뎅,,,