import pandas as pd
import numpy as np

# データ読み込み
df = pd.read_csv('data/raw/SSDSE-B-2026.csv', encoding='utf-8', skiprows=1)
print(df.shape)
print(df.dtypes)
print(df.describe())

# 「レーダーチャート」の文脈で扱う場合の例：
# 分野: 可視化
# 関連手法は同カテゴリの他用語を参照してください。

# matplotlib で 47 都道府県の主要 5 指標をレーダーチャート（東京vs沖縄）
import pandas as pd, numpy as np, matplotlib.pyplot as plt
df = pd.read_csv('data/raw/SSDSE-B-2026.csv', encoding='utf-8', skiprows=1)
df['高齢化率'] = df['A1301'] / df['A1101'] * 100
cols = ['高齢化率','A1101','A4101','C3301','E1101']
labels = ['高齢化率','総人口','出生数','製造業出荷','学校数']
metrics = df.set_index('Prefecture')[cols]
norm = (metrics - metrics.min()) / (metrics.max() - metrics.min())

angles = np.linspace(0, 2*np.pi, len(cols), endpoint=False).tolist()
angles += angles[:1]
fig, ax = plt.subplots(figsize=(6,6), subplot_kw=dict(polar=True))
for pref in ['東京都','沖縄県']:
    vals = norm.loc[pref].tolist(); vals += vals[:1]
    ax.plot(angles, vals, label=pref)
    ax.fill(angles, vals, alpha=0.2)
ax.set_xticks(angles[:-1]); ax.set_xticklabels(labels)
ax.legend(loc='upper right'); plt.tight_layout(); plt.savefig('radar.png', dpi=150)

# plotly Express で同様
import plotly.express as px
long = norm.loc[['東京都','沖縄県']].reset_index().melt(id_vars='Prefecture')
long.columns = ['Prefecture','metric','value']
fig = px.line_polar(long, r='value', theta='metric',
                    color='Prefecture', line_close=True)
fig.write_html('radar.html')

# 正規化方法の選択 : min-max vs standardization
from sklearn.preprocessing import StandardScaler
norm2 = pd.DataFrame(StandardScaler().fit_transform(metrics),
                      index=metrics.index, columns=metrics.columns)
print(norm2.loc['東京都'])

# レーダーの面積で「指標豊かさ」を比較
areas = []
for pref in metrics.index:
    v = norm.loc[pref].values
    area = 0.5 * np.sum(v * np.roll(v, -1) * np.sin(2*np.pi/len(v)))
    areas.append((pref, area))
top3 = sorted(areas, key=lambda x: -x[1])[:3]
print('面積トップ3:', top3)

import pandas as pd
df = pd.read_csv('data/raw/SSDSE-B-2026.csv', encoding='utf-8', skiprows=1)
df['高齢化率'] = df['A1301']/df['A1101']*100
cols = ['高齢化率','A1101','A4101']
raw = df.set_index('Prefecture')[cols].loc[['東京都','沖縄県']]
print(raw)  # スケール差で潰れる例
norm = (raw - raw.min()) / (raw.max() - raw.min())
print(norm)

import numpy as np, matplotlib.pyplot as plt
labels = ['高齢化率','出生率','学校数','医療','製造']
vals_a = [0.9, 0.2, 0.7, 0.8, 0.4]; vals_b = [0.3, 0.8, 0.5, 0.6, 0.9]
angles = np.linspace(0,2*np.pi,len(labels),endpoint=False); angles = np.append(angles, angles[0])
vals_a += vals_a[:1]; vals_b += vals_b[:1]
fig, ax = plt.subplots(subplot_kw=dict(polar=True))
ax.plot(angles, vals_a); ax.plot(angles, vals_b)
ax.set_xticks(angles[:-1]); ax.set_xticklabels(labels)
plt.tight_layout(); plt.savefig('compare.png', dpi=150)

labels = [f'指標{i}' for i in range(10)]
if len(labels) > 8: print('WARNING: 8 指標を超えたら平行座標を検討')