실습 6 — 기준선, 직선 모델과 견주기

import numpy as np
import pandas as pd
import torch
from torch import nn
from torch.utils.data import TensorDataset, DataLoader
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.metrics import mean_absolute_error, root_mean_squared_error

df = pd.read_csv("daily_orders.csv")
df["일차"] = range(len(df))
features = ["일차", "주말", "공휴일", "기온", "강수mm", "광고비만원", "할인행사"]
X_train, X_test, y_train, y_test = train_test_split(
    df[features], df["주문량"], test_size=0.2, random_state=42
)
scaler = StandardScaler().fit(X_train)
X_train_s, X_test_s = scaler.transform(X_train), scaler.transform(X_test)
print(X_train_s.shape, X_test_s.shape)

from sklearn.linear_model import LinearRegression

# 기준선: 학습용 주문량의 평균을 언제나 답한다
base_pred = np.full(len(y_test), y_train.mean())

# 직선 모델: 특성마다 가중치 하나씩. 은닉층 없는 신경망(뉴런 하나)과 같은 모양이다
lin = LinearRegression().fit(X_train_s, y_train)
lin_pred = lin.predict(X_test_s)
print("준비 완료")

# 실습 4와 똑같은 신경망과 학습(같은 씨앗이라 같은 결과가 나온다)
torch.manual_seed(42)
model = nn.Sequential(nn.Linear(7, 32), nn.ReLU(), nn.Linear(32, 16), nn.ReLU(), nn.Linear(16, 1))
opt = torch.optim.Adam(model.parameters(), lr=0.01)
loss_fn = nn.MSELoss()
X_train_t = torch.tensor(X_train_s, dtype=torch.float32)
y_train_t = torch.tensor(y_train.to_numpy() / 100, dtype=torch.float32).view(-1, 1)
loader = DataLoader(TensorDataset(X_train_t, y_train_t), batch_size=32, shuffle=True,
                    generator=torch.Generator().manual_seed(42))
for epoch in range(200):
    for xb, yb in loader:
        opt.zero_grad()
        loss = loss_fn(model(xb), yb)
        loss.backward()
        opt.step()

with torch.no_grad():
    nn_pred = model(torch.tensor(X_test_s, dtype=torch.float32)).numpy().ravel() * 100
print("학습 완료")

# 세 모델의 MAE 와 RMSE 를 표 하나로 모은다
rows = []
for name, pred in [("기준선(평균)", base_pred), ("직선 모델", lin_pred), ("신경망", nn_pred)]:
    rows.append({"모델": name,
                 "MAE": round(mean_absolute_error(y_test, pred), 1),
                 "RMSE": round(root_mean_squared_error(y_test, pred), 1)})
print(pd.DataFrame(rows).to_string(index=False))

# 주말 할인행사 날만 골라 두 모델이 얼마나 빗나가는지 견준다
mask = ((X_test["주말"] == 1) & (X_test["할인행사"] == 1)).to_numpy()
print("주말 할인행사 날", mask.sum(), "일")
print("  직선 모델 MAE:", round(mean_absolute_error(y_test[mask], lin_pred[mask]), 1))
print("  신경망 MAE:", round(mean_absolute_error(y_test[mask], nn_pred[mask]), 1))