머신 러닝 (5) EDA

차트 작성 시 한글 깨짐 방지를 위한 koreanize-matplotlib 설치

 !pip install koreanize-matplotlib

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

import koreanize_matplotlib

import seaborn as sns

from sklearn.datasets import load_diabetes

from scipy import stats

import warnings

warnings.filterwarnings('ignore')

pima_columns = ['pregnancies', 'glucose', 'blood_pressure', 'skin_thickness', 'insulin', 'bmi', 'diabetes_pedigree_function', 'age', 'outcome']

pima_data_url = 'https://raw.githubusercontent.com/jbrownlee/Datasets/master/pima-indians-diabetes.data.csv'

df = pd.read_csv(pima_data_url, names=pima_columns)

df.shape

df.info()

# 1. 결측치(NaN) 확인

missing_values = df.isnull().sum()

display(missing_values)

# 2. 숨겨진 결측치 (0값) 탐색

zero_cols = ['glucose', 'blood_pressure', 'skin_thickness', 'insulin', 'bmi']

for col in zero_cols:

    zero_count = (df[col] == 0).sum()

    print(f" - {col}: {zero_count}개 ({zero_count/len(df)*100:.2f}%)")

# 수치형 데이터 요약

df.describe()

df.hist(figsize=(12, 10), bins=20, edgecolor='black')

plt.suptitle("전체 변수 히스토그램", fontsize=16)

plt.tight_layout()

plt.show()

# 빈도 및 비율 확인

outcome_counts = df['outcome'].value_counts()

outcome_ratios = df['outcome'].value_counts(normalize=True)

print(f"0 (정상): {outcome_counts[0]}명 ({outcome_ratios[0]*100:.1f}%)")

print(f"1 (당뇨): {outcome_counts[1]}명 ({outcome_ratios[1]*100:.1f}%)")

# 시각화 (Count Plot)

plt.figure(figsize=(2, 4))

sns.countplot(x='outcome', data=df, palette='pastel')

plt.title('Distribution of Outcome (Target Variable)')

plt.xlabel('Outcome (0: Normal, 1: Diabetes)')

plt.ylabel('Count')

plt.show()

# Box Plot 시각화

plt.figure(figsize=(6, 3))

sns.boxplot(x=df['blood_pressure'], color='lightskyblue')

plt.title('Box Plot of Blood Pressure')

plt.show()

# IQR 계산 로직

Q1 = df['blood_pressure'].quantile(0.25)

Q3 = df['blood_pressure'].quantile(0.75)

IQR = Q3 - Q1

lower_bound = Q1 - 1.5 * IQR

upper_bound = Q3 + 1.5 * IQR

print(f"IQR: {IQR}")

print(f"이상치 하한 경계: {lower_bound}")

print(f"이상치 상한 경계: {upper_bound}")

# 상관계수 행렬 계산

corr_matrix = df.corr()

# 히트맵 시각화

plt.figure(figsize=(8, 6))

sns.heatmap(corr_matrix, annot=True, fmt='.2f', cmap='coolwarm', vmin=-1, vmax=1)

plt.title('Correlation Heatmap of Variables')

plt.show()

plt.figure(figsize=(6, 4))

# 변경된 컬럼명 glucose, outcome 적용

sns.kdeplot(data=df[df['outcome'] == 0]['glucose'], label='Non-Diabetic (0)', fill=True, color='blue', alpha=0.3)

sns.kdeplot(data=df[df['outcome'] == 1]['glucose'], label='Diabetic (1)', fill=True, color='orange', alpha=0.3)

plt.title('Glucose Distribution by Outcome')

plt.xlabel('Glucose')

plt.ylabel('Density')

plt.legend()

plt.show()

fig, axes = plt.subplots(1, 2, figsize=(8, 4))

# Box Plot

sns.boxplot(x='outcome', y='age', data=df, ax=axes[0], palette='Set2')

axes[0].set_title('Box Plot: Age vs Outcome')

# Violin Plot

sns.violinplot(x='outcome', y='age', data=df, ax=axes[1], palette='Set2')

axes[1].set_title('Violin Plot: Age vs Outcome')

plt.tight_layout()

plt.show()