Spaces:
Build error
Build error
File size: 9,798 Bytes
fab9847 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 | """
run.py — 主入口:数据更新 → 引擎 → 分析 → 对冲 → 因果 → Dashboard
===================================================================
用法:
python run.py # 完整流程(含数据更新+新闻)
python run.py --skip-update # 跳过数据更新,直接用现有数据
"""
import os, json, webbrowser, argparse
import pandas as pd
from config import BASE_DIR, OUTPUT_DIR, OUTPUT_FILES, INDUSTRIES, PRICE_COLS
from core.engine import load_panel, run_walk_forward
from core.analysis import apply_industry_rules, generate_all_reports, evaluate_results, run_ablation
from core.hedging import compute_all_industry_hedges, backtest_hedging
from core.feature_selection import run_feature_funnel
os.chdir(BASE_DIR)
os.makedirs(OUTPUT_DIR, exist_ok=True)
def run_benchmark(panel_path, benchmark, price_col, features_override=None):
"""对单个油价基准运行完整 walk-forward 流程。"""
print(f"\n{'─'*65}")
print(f" Benchmark: {benchmark} ({price_col})")
print(f"{'─'*65}")
# Load panel
panel, features = load_panel(panel_path, price_col=price_col)
# Walk-forward
results, shap_data = run_walk_forward(panel, features)
print(f" 测试月数: {len(results)}")
# Industry rules
for i, row in results.iterrows():
rules = apply_industry_rules(row)
for k, v in rules.items():
results.at[i, k] = v
# NLG reports
reports = generate_all_reports(results)
# Tag benchmark
results['benchmark'] = benchmark
return results, shap_data, reports, panel, features
def main(skip_update=False):
# ════ Step 0: Data Update ════
if not skip_update:
print("═" * 65)
print("Step 0: 全特征 API 数据更新")
print("═" * 65)
try:
from pipeline.live_data import main as live_main
live_main()
panel_path = 'output/panel_monthly_live.csv'
if os.path.exists(panel_path):
print(f"✓ 使用更新后的面板: {panel_path}")
else:
panel_path = 'output/panel_monthly.csv'
except Exception as e:
print(f"⚠ 数据更新跳过: {e}")
panel_path = 'output/panel_monthly.csv'
else:
print("跳过数据更新")
if os.path.exists('output/panel_monthly_live.csv'):
panel_path = 'output/panel_monthly_live.csv'
else:
panel_path = 'output/panel_monthly.csv'
# ════ Step 1: Feature Selection Funnel ════
print("\n" + "═" * 65)
print("Step 1: 特征筛选漏斗 (329→17)")
print("═" * 65)
funnel = run_feature_funnel('output/panel_monthly.csv')
with open(OUTPUT_FILES['feat_sel'], 'w', encoding='utf-8') as f:
json.dump(funnel, f, ensure_ascii=False, indent=2)
print(f"✓ 特征筛选: {OUTPUT_FILES['feat_sel']}")
# ════ Step 2: Walk-Forward for EACH benchmark ════
print("\n" + "═" * 65)
print("Step 2: Walk-Forward 预测 (WTI + Brent)")
print("═" * 65)
all_results = {}
all_shap = {}
all_reports = {}
all_panels = {}
all_features = {}
for benchmark, price_col in PRICE_COLS.items():
results, shap_data, reports, panel, features = run_benchmark(
panel_path, benchmark, price_col)
all_results[benchmark] = results
all_shap[benchmark] = shap_data
all_reports[benchmark] = reports
all_panels[benchmark] = panel
all_features[benchmark] = features
# Use WTI as primary for hedging/evaluation (backward compat)
primary = 'WTI'
results = all_results[primary]
# ════ Step 3: Hedging (based on WTI) ════
print("\n" + "═" * 65)
print("Step 3: 对冲决策计算")
print("═" * 65)
latest = results.iloc[-1]
hedging_data = compute_all_industry_hedges(latest)
hedging_json = {}
for ind, hd in hedging_data.items():
hedging_json[ind] = {
'industry_zh': hd['industry_zh'],
'exposure': hd['exposure'],
'elasticity': hd['elasticity'],
'recommended_ratio': hd['recommended_ratio'],
'recommended_ratio_pct': hd['recommended_ratio_pct'],
'recommended_tool': hd['recommended_tool'],
'rationale': hd['rationale'],
'matrix': hd['matrix'],
'tool_comparison': hd['tool_comparison'],
}
print(f" {hd['industry_zh']}: 推荐对冲 {hd['recommended_ratio_pct']}, "
f"工具={hd['recommended_tool']}")
with open(OUTPUT_FILES['hedging'], 'w', encoding='utf-8') as f:
json.dump(hedging_json, f, ensure_ascii=False, indent=2)
print(f"✓ Hedging: {OUTPUT_FILES['hedging']}")
# Hedge Backtest
print(" [回测对冲策略 — 过去60月]")
backtest_data = backtest_hedging(results)
backtest_json = {}
for ind, bt in backtest_data.items():
backtest_json[ind] = bt
print(f" {bt['industry_zh']}: 累计节省${bt['total_saving']:.1f}M, "
f"波动率降低{bt['vol_reduction']:.0f}%, "
f"最大回撤改善${bt['dd_improvement']:.1f}M")
with open(OUTPUT_FILES['backtest'], 'w', encoding='utf-8') as f:
json.dump(backtest_json, f, ensure_ascii=False, indent=2)
print(f"✓ Backtest: {OUTPUT_FILES['backtest']}")
# ════ Step 4: NLG Reports ════
print("\n" + "═" * 65)
print("Step 4: NLG 报告生成")
print("═" * 65)
for bm, reports in all_reports.items():
print(f" {bm}: {len(reports)} 份报告")
# ════ Step 5: Evaluation ════
for bm, res in all_results.items():
print(f"\n--- Evaluation: {bm} ---")
evaluate_results(res)
# ════ Step 6: Save ════
print("\n" + "═" * 65)
print("Step 6: 保存结果")
print("═" * 65)
# Save per-benchmark results
for bm, res in all_results.items():
out_path = os.path.join(OUTPUT_DIR, f'v2_results_{bm}.csv')
res.to_csv(out_path, index=False)
print(f"✓ 结果 [{bm}]: {out_path}")
# Also save primary as the main results (backward compat)
results.to_csv(OUTPUT_FILES['results'], index=False)
print(f"✓ 结果 [primary]: {OUTPUT_FILES['results']}")
# SHAP (primary)
with open(OUTPUT_FILES['shap'], 'w', encoding='utf-8') as f:
json.dump(all_shap[primary][-12:], f, ensure_ascii=False, indent=2)
print(f"✓ SHAP: {OUTPUT_FILES['shap']}")
# NLG (merge all benchmarks)
merged_reports = {}
for bm, reps in all_reports.items():
for dt, report in reps.items():
key = f"{dt}_{bm}" if bm != primary else dt
merged_reports[key] = report
# Also save per-benchmark
with open(os.path.join(OUTPUT_DIR, f'v2_nlg_{bm}.json'), 'w', encoding='utf-8') as f:
json.dump(reps, f, ensure_ascii=False, indent=2)
with open(OUTPUT_FILES['nlg'], 'w', encoding='utf-8') as f:
json.dump(merged_reports, f, ensure_ascii=False, indent=2)
print(f"✓ NLG: {OUTPUT_FILES['nlg']}")
# Scenarios (primary)
scenario_data = {}
for _, row in results.tail(12).iterrows():
dt = pd.Timestamp(row['test_date']).strftime('%Y-%m')
scenario_data[dt] = {
'base': round(row['scenario_base'] * 100, 2),
'vix_shock': round(row['scenario_vix_shock'] * 100, 2),
'supply_cut': round(row['scenario_supply_cut'] * 100, 2),
'demand_crash': round(row['scenario_demand_crash'] * 100, 2),
}
with open(OUTPUT_FILES['scenarios'], 'w', encoding='utf-8') as f:
json.dump(scenario_data, f, indent=2)
print(f"✓ Scenarios: {OUTPUT_FILES['scenarios']}")
# Regime (primary)
regime_data = {}
for _, row in results.iterrows():
dt = pd.Timestamp(row['test_date']).strftime('%Y-%m')
regime_data[dt] = {
'match': row.get('regime_match', 'Unknown'),
'similarity': row.get('regime_similarity', 0),
'type': row.get('regime_type', 'normal'),
}
with open(OUTPUT_FILES['regime'], 'w', encoding='utf-8') as f:
json.dump(regime_data, f, ensure_ascii=False, indent=2)
print(f"✓ Regime: {OUTPUT_FILES['regime']}")
# ════ Step 7: Ablation (primary only) ════
print("\n" + "═" * 65)
print("Step 7: 消融实验")
print("═" * 65)
ablation_results = run_ablation(all_panels[primary], all_features[primary])
with open(OUTPUT_FILES['ablation'], 'w') as f:
json.dump(ablation_results, f, indent=2)
print(f"✓ Ablation: {OUTPUT_FILES['ablation']}")
# ════ Step 7b: Causal Analysis ════
print("\n" + "═" * 65)
print("Step 7b: 因果因子网络分析")
print("═" * 65)
try:
from pipeline.causal_analysis import run_full_causal_analysis
causal_result = run_full_causal_analysis(panel_path)
print(f"✓ 因果分析: {OUTPUT_FILES.get('causal', 'output/causal_analysis.json')}")
except Exception as e:
print(f"⚠ 因果分析跳过: {e}")
# ════ Step 8: Done ════
print("\n" + "═" * 65)
print("✅ 全部完成!")
print("═" * 65)
print(" 启动前端: cd frontend && npm run dev")
print(" 启动API: python api_server.py")
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='油价风险分析平台 — 一键启动')
parser.add_argument('--skip-update', action='store_true',
help='跳过 FRED/EIA 数据更新,直接使用现有数据')
args = parser.parse_args()
main(skip_update=args.skip_update)
|