시계열 매출 예측 분석: Prophet과 ARIMA로 비즈니스 인사이트 도출하기

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from statsmodels.tsa.seasonal import seasonal_decompose
from statsmodels.tsa.stattools import adfuller
import warnings
warnings.filterwarnings('ignore')

# 데이터 로드
df = pd.read_csv('train.csv', parse_dates=['date'])
print(df.head())
print(f"기간: {df['date'].min()} ~ {df['date'].max()}")
print(f"총 레코드: {len(df):,}건")

# 특정 아이템 필터링
item_sales = df[(df['store']==1) & (df['item']==1)].copy()
item_sales.set_index('date', inplace=True)

# 시계열 플롯
plt.figure(figsize=(14, 6))
plt.plot(item_sales.index, item_sales['sales'], linewidth=0.8)
plt.title('Store 1, Item 1 일별 판매량', fontsize=14)
plt.xlabel('날짜')
plt.ylabel('판매량')
plt.grid(alpha=0.3)
plt.tight_layout()
plt.show()

decomposition = seasonal_decompose(item_sales['sales'], model='additive', period=365)

fig, axes = plt.subplots(4, 1, figsize=(14, 10))
decomposition.observed.plot(ax=axes[0], title='Original')
decomposition.trend.plot(ax=axes[1], title='Trend')
decomposition.seasonal.plot(ax=axes[2], title='Seasonality')
decomposition.resid.plot(ax=axes[3], title='Residuals')
plt.tight_layout()
plt.show()

result = adfuller(item_sales['sales'])
print(f'ADF Statistic: {result[0]:.4f}')
print(f'p-value: {result[1]:.4f}')

if result[1] > 0.05:
    print("→ 비정상 시계열입니다. 차분이 필요합니다.")
else:
    print("→ 정상 시계열입니다.")

from statsmodels.tsa.arima.model import ARIMA

# Train/Test 분할 (마지막 90일을 테스트셋으로)
train = item_sales[:-90]
test = item_sales[-90:]

# ARIMA(1,1,1) 모델 학습
model = ARIMA(train['sales'], order=(1,1,1))
model_fit = model.fit()
print(model_fit.summary())

# 예측
forecast = model_fit.forecast(steps=90)

# 시각화
plt.figure(figsize=(14, 6))
plt.plot(train.index, train['sales'], label='Train', color='blue')
plt.plot(test.index, test['sales'], label='Test (Actual)', color='green')
plt.plot(test.index, forecast, label='ARIMA Forecast', color='red', linestyle='--')
plt.legend()
plt.title('ARIMA 예측 결과')
plt.tight_layout()
plt.show()

from prophet import Prophet

# Prophet 입력 형식 변환 (ds, y 컬럼 필요)
prophet_df = item_sales.reset_index()[['date', 'sales']]
prophet_df.columns = ['ds', 'y']

train_prophet = prophet_df[:-90]
test_prophet = prophet_df[-90:]

# 모델 학습
model = Prophet(
    yearly_seasonality=True,
    weekly_seasonality=True,
    daily_seasonality=False,
    seasonality_mode='multiplicative'
)
model.fit(train_prophet)

# 미래 데이터프레임 생성 및 예측
future = model.make_future_dataframe(periods=90)
forecast = model.predict(future)

# 예측 결과 시각화
fig = model.plot(forecast)
plt.title('Prophet 예측 결과')
plt.tight_layout()
plt.show()

# 컴포넌트 분해
fig2 = model.plot_components(forecast)
plt.tight_layout()
plt.show()

from sklearn.metrics import mean_squared_error, mean_absolute_percentage_error

# ARIMA
arima_rmse = np.sqrt(mean_squared_error(test['sales'], forecast))
arima_mape = mean_absolute_percentage_error(test['sales'], forecast) * 100

# Prophet
prophet_pred = forecast.iloc[-90:]['yhat'].values
prophet_rmse = np.sqrt(mean_squared_error(test_prophet['y'], prophet_pred))
prophet_mape = mean_absolute_percentage_error(test_prophet['y'], prophet_pred) * 100

results = pd.DataFrame({
    'Model': ['ARIMA', 'Prophet'],
    'RMSE': [arima_rmse, prophet_rmse],
    'MAPE (%)': [arima_mape, prophet_mape]
})

print(results)