import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import OneHotEncoder, StandardScaler
from sklearn.compose import ColumnTransformer
from sklearn.pipeline import Pipeline
from sklearn.ensemble import RandomForestClassifier, StackingClassifier, GradientBoostingClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, classification_report, roc_curve, roc_auc_score
import xgboost as xgb
import lightgbm as lgb
import joblib
import os
from datetime import datetime
# 1. 데이터 로드
train = pd.read_csv("train.csv")
test = pd.read_csv("test.csv")
# 2. 주요 변수 선택 (예시)
main_features = [
'Pclass', 'Sex', 'Age', 'SibSp', 'Parch', 'Fare', 'Embarked'
]
# 3. 피처/타겟 분리
X = train[main_features]
y = train['Survived']
X_test = test[main_features]
# 4. 훈련/검증 분리
X_train, X_val, y_train, y_val = train_test_split(
X, y, test_size=0.2, random_state=42, stratify=y
)
# 5. 전처리 파이프라인 (결측치 처리 포함)
numeric_feats = X_train.select_dtypes(include=['int64', 'float64']).columns.tolist()
categorical_feats = X_train.select_dtypes(include=['object']).columns.tolist()
numeric_transformer = Pipeline([
('imputer', SimpleImputer(strategy='median')),
('scaler', StandardScaler())
])
categorical_transformer = Pipeline([
('imputer', SimpleImputer(strategy='most_frequent')),
('encoder', OneHotEncoder(handle_unknown='ignore', sparse_output=False))
])
preprocessor = ColumnTransformer([
('num', numeric_transformer, numeric_feats),
('cat', categorical_transformer, categorical_feats)
])
# 6. Base 모델 정의
base_estimators = [
('rf', RandomForestClassifier(n_estimators=100, max_depth=5, random_state=42)),
('xgb', xgb.XGBClassifier(n_estimators=100, max_depth=3, tree_method='hist', random_state=42, use_label_encoder=False, eval_metric='logloss')),
('lgb', lgb.LGBMClassifier(n_estimators=100, max_depth=3, random_state=42))
]
# 7. Stacking 앙상블 (파이프라인 포함)
stacking = StackingClassifier(
estimators=[(name, Pipeline([('pre', preprocessor), (name, model)])) for name, model in base_estimators],
final_estimator=GradientBoostingClassifier(n_estimators=100, random_state=42),
cv=5,
n_jobs=-1,
passthrough=False
)
# 8. 학습
stacking.fit(X_train, y_train)
# 9. 검증 예측 및 평가
val_preds = stacking.predict(X_val)
val_probs = stacking.predict_proba(X_val)[:, 1]
print("검증 정확도:", accuracy_score(y_val, val_preds))
print(classification_report(y_val, val_preds))
# 10. ROC Curve (fig, ax)
fpr, tpr, _ = roc_curve(y_val, val_probs)
auc = roc_auc_score(y_val, val_probs)
fig, ax = plt.subplots(figsize=(6, 5))
ax.plot(fpr, tpr, label=f'Stacking (AUC={auc:.3f})')
ax.plot([0, 1], [0, 1], 'k--', label='Random')
ax.set_xlabel('False Positive Rate')
ax.set_ylabel('True Positive Rate')
ax.set_title('ROC Curve - Titanic Survival (Stacking)')
ax.legend()
ax.grid(True)
plt.tight_layout()
plt.show()
# 11. Test 데이터 예측 및 제출 파일 생성
test_preds = stacking.predict(X_test)
submission = pd.DataFrame({
'PassengerId': test['PassengerId'],
'Survived': test_preds
})
# 12. 모델 내보내기
model_path = "titanic_stacking_model.pkl"
os.makedirs("models", exist_ok=True)
with open(model_path, "wb") as f:
joblib.dump(stacking, f)
print(f"모델이 {model_path}에 저장되었습니다.")
import streamlit as st
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import joblib
## 페이지 설정
st.set_page_config(
page_title='Titanic Survival Prediction',
layout='wide'
)
st.title("Titanic Main page, Dashboard")
## 데이터 불러오기
@st.cache_data # 데이터 불러올 땐 무조건 있어야 함
def load_data():
train = pd.read_csv('train.csv')
# 추가로 이곳에서 데이터 처리 (불필요한 컬럼 제거 등)을 할 수 있음
# 추가로 local에서가 아닌 sql에서 수집하는 방법 작성
return train
train = load_data()
# st.dataframe(train) # 확인용
## 탭 생성
tab1, tab2, tab3 = st.tabs(['EDA', 'Statistics', 'Prediction'])
# EDA 탭
with tab1:
st.header("Exploratory Data Analysis")
# 데이터 미리보기
st.subheader("Data Preview")
st.dataframe(train.head())
# 생존자 분포
st.subheader("Survival Distribution")
fig, ax = plt.subplots(figsize=(8, 6))
sns.countplot(data=train, x='Survived')
plt.title("Survival Count")
st.pyplot(fig)
# 성별별 생존율
st.subheader("Survival Rate by Gender")
fig, ax = plt.subplots(figsize=(8, 6))
sns.barplot(data=train, x='Sex', y='Survived')
plt.title("Survival Rate by Gender")
st.pyplot(fig)
# 객실 등급별 생존율
st.subheader("Survival Rate by Passenger Class")
fig, ax = plt.subplots(figsize=(8, 6))
sns.barplot(data=train, x='Pclass', y='Survived')
plt.title("Survival Rate by Passenger Class")
st.pyplot(fig)
# Statistics 탭
with tab2:
st.header("Statistical Analysis")
# 기본 통계량
st.subheader("Basic Statistics")
st.dataframe(train.describe())
# 상관관계 분석
st.subheader("Correlation Analysis")
numeric_cols = train.select_dtypes(include=['int64', 'float64']).columns
corr = train[numeric_cols].corr()
fig, ax = plt.subplots(figsize=(10, 8))
sns.heatmap(corr, annot=True, cmap='coolwarm', ax=ax)
st.pyplot(fig)
# 결측치 분석
st.subheader("Missing Values Analysis")
missing_data = pd.DataFrame({
'Missing Values': train.isnull().sum(),
'Percentage': (train.isnull().sum() / len(train)) * 100
})
st.dataframe(missing_data)
with tab3:
st.header('예측')
@st.cache_resource # 세션과 관련 있는 데코레이터
def load_model():
return joblib.load('titanic_stacking_model.pkl')
model = load_model()
# st.write(model)
## 입력 폼
st.subheader('고객 정보 입력')
# col1, col2 = st.columns(2)
# # 입력값 기준점은 train.csv 파일 기준에 작성
# with col1:
# pclass = st.selectbox('Passenger Class', [1,2,3])
# with col2:
# fare = st.number_input('Fare', min_value=0.0, max_value=500.0, value=50.0)
# if st.button('Predict Survival'):
# input_data = pd.DataFrame({
# 'Pclass': [pclass],
# 'Fare': [fare]
# })
# st.dataframe(input_data)
# model.predict(input_data)
col1, col2 = st.columns(2)
with col1:
pclass = st.selectbox("Passenger Class", [1, 2, 3])
sex = st.selectbox("Sex", ["male", "female"])
age = st.number_input("Age", min_value=0, max_value=100, value=30)
with col2:
sibsp = st.number_input("Number of Siblings/Spouses", min_value=0, max_value=10, value=0)
parch = st.number_input("Number of Parents/Children", min_value=0, max_value=10, value=0)
fare = st.number_input("Fare", min_value=0.0, max_value=500.0, value=50.0)
embarked = st.selectbox("Port of Embarkation", ["C", "Q", "S"])
# 예측 버튼
if st.button("Predict Survival"):
# 입력 데이터 준비
input_data = pd.DataFrame({
'Pclass': [pclass],
'Sex': [sex],
'Age': [age],
'SibSp': [sibsp],
'Parch': [parch],
'Fare': [fare],
'Embarked': [embarked]
})
# 예측
prediction = model.predict(input_data)[0]
probability = model.predict_proba(input_data)[0][1]
# 결과 표시
st.subheader("Prediction Result")
if prediction == 1:
st.success(f"Survival Probability: {probability:.2%}")
st.write("The passenger is predicted to survive.")
else:
st.error(f"Survival Probability: {probability:.2%}")
st.write("The passenger is predicted to not survive.")