from sklearn.metrics import mean_squared_error, mean_absolute_error
import numpy as np

y_true = np.array([10, 12, 8, 15, 30])
y_pred = np.array([9, 11, 10, 14, 12])

print('MSE:', mean_squared_error(y_true, y_pred))
print('MAE:', mean_absolute_error(y_true, y_pred))
print('RMSE:', np.sqrt(mean_squared_error(y_true, y_pred)))

import pandas as pd
import numpy as np

# データ読み込み（SSDSE-B 都道府県・47 県 × 約 112 列）
df = pd.read_csv('data/raw/SSDSE-B-2026.csv', encoding='cp932', skiprows=1)
df = df[df['年度'] == 2023].reset_index(drop=True)
print('shape:', df.shape)
print('列の先頭:', df.columns.tolist()[:6])

# 必要な列だけ取り出して整形
features = ['総人口', '15歳未満人口', '65歳以上人口', '県内総生産']
df_use = df[features].copy()
print(df_use.describe())

from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_squared_error, r2_score

X = df[['総人口', '65歳以上人口', '県内総生産']].fillna(0).values
y = df['合計特殊出生率'].fillna(df['合計特殊出生率'].median()).values

X_tr, X_te, y_tr, y_te = train_test_split(X, y, test_size=0.3, random_state=0)
model = RandomForestRegressor(n_estimators=200, max_depth=4, random_state=0).fit(X_tr, y_tr)

pred_tr = model.predict(X_tr)
pred_te = model.predict(X_te)
print(f'train R^2 = {r2_score(y_tr, pred_tr):.3f}')
print(f'test  R^2 = {r2_score(y_te, pred_te):.3f}')
print(f'test RMSE = {np.sqrt(mean_squared_error(y_te, pred_te)):.4f}')

import matplotlib.pyplot as plt

plt.figure(figsize=(7,5))
plt.scatter(y_te, pred_te, alpha=0.7, edgecolor='k')
lims = [min(y_te.min(), pred_te.min()), max(y_te.max(), pred_te.max())]
plt.plot(lims, lims, 'r--', linewidth=2, label='完全予測ライン')
plt.xlabel('実測 出生率')
plt.ylabel('予測 出生率')
plt.title('損失関数 を使ったモデルの予測精度（SSDSE-B-2026）')
plt.legend()
plt.tight_layout()
plt.savefig('out_loss-function.png', dpi=150)

from sklearn.model_selection import cross_val_score

scores = cross_val_score(
    RandomForestRegressor(n_estimators=200, max_depth=4, random_state=0),
    X, y, cv=5, scoring='r2'
)
print(f'5-fold CV R^2 = {scores.mean():.3f} (±{scores.std():.3f})')
print('各 fold:', np.round(scores, 3))

import pandas as pd
df = pd.read_csv('data/raw/SSDSE-B-2026.csv', encoding='cp932', skiprows=1)
df = df[df['年度'] == 2023].reset_index(drop=True)
print(df.shape, df.head(3))

from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import Ridge

pipe = Pipeline([
    ('scaler', StandardScaler()),
    ('model',  Ridge(alpha=1.0)),
])
pipe.fit(X_tr, y_tr)
print('R^2 =', pipe.score(X_te, y_te))

from sklearn.model_selection import GridSearchCV

params = {'model__alpha': [0.01, 0.1, 1.0, 10.0, 100.0]}
gs = GridSearchCV(pipe, params, cv=5, scoring='r2', n_jobs=-1)
gs.fit(X, y)
print('best:', gs.best_params_, 'score:', gs.best_score_)

import matplotlib.pyplot as plt
import json

pred = gs.predict(X_te)
plt.figure(figsize=(7,5))
plt.scatter(y_te, pred, alpha=0.7, edgecolor='k')
plt.plot([y_te.min(), y_te.max()], [y_te.min(), y_te.max()], 'r--')
plt.xlabel('実測'); plt.ylabel('予測'); plt.title('損失関数 結果')
plt.tight_layout(); plt.savefig('result_loss-function.png', dpi=150)

with open('result_loss-function.json', 'w', encoding='utf-8') as f:
    json.dump({'best_params': gs.best_params_,
               'cv_score': gs.best_score_,
               'test_score': gs.score(X_te, y_te)}, f, ensure_ascii=False, indent=2)