|
|
@@ -0,0 +1,195 @@
|
|
|
+#!/usr/bin/env python3
|
|
|
+# -*- coding: utf-8 -*-
|
|
|
+"""
|
|
|
+趋势质量评估器 - 参数优化测试 (简化版)
|
|
|
+测试不同参数组合的回测表现
|
|
|
+"""
|
|
|
+
|
|
|
+import sys
|
|
|
+sys.path.insert(0, '/root/.openclaw/workspace/trend-quality-evaluator')
|
|
|
+
|
|
|
+import numpy as np
|
|
|
+import pandas as pd
|
|
|
+from trend_quality_evaluator import fetch_stock_data
|
|
|
+import warnings
|
|
|
+warnings.filterwarnings('ignore')
|
|
|
+
|
|
|
+print("="*70)
|
|
|
+print("趋势质量评估器 - 参数优化测试")
|
|
|
+print("="*70)
|
|
|
+
|
|
|
+# 获取数据
|
|
|
+df = fetch_stock_data("399673", "2019-01-01", "2026-03-06", "d")
|
|
|
+if df is None:
|
|
|
+ print("数据获取失败")
|
|
|
+ exit(1)
|
|
|
+
|
|
|
+print(f"\n✓ 数据获取成功: {len(df)}条")
|
|
|
+
|
|
|
+# 预计算指标
|
|
|
+def calc_indicators(data):
|
|
|
+ d = data.copy()
|
|
|
+ high, low, close = d['high'], d['low'], d['close']
|
|
|
+
|
|
|
+ # ADX
|
|
|
+ plus_dm = high.diff().where(high.diff() > 0, 0)
|
|
|
+ minus_dm = low.diff().abs().where(low.diff().abs() > 0, 0)
|
|
|
+ tr = pd.concat([high-low, (high-close.shift()).abs(), (low-close.shift()).abs()], axis=1).max(axis=1)
|
|
|
+ atr = tr.rolling(14).mean()
|
|
|
+ dx = (plus_dm.rolling(14).mean() / atr * 100).abs()
|
|
|
+ d['adx'] = dx.rolling(14).mean()
|
|
|
+
|
|
|
+ # MA斜率
|
|
|
+ d['ma20'] = d['close'].rolling(20).mean()
|
|
|
+ d['ma_slope'] = d['ma20'] / d['ma20'].shift(5)
|
|
|
+
|
|
|
+ # ATR比率
|
|
|
+ d['atr_ratio'] = tr.rolling(14).mean() / tr.rolling(50).mean()
|
|
|
+
|
|
|
+ # 成交量
|
|
|
+ d['vol_ratio'] = d['volume'] / d['volume'].rolling(20).mean()
|
|
|
+
|
|
|
+ d['above_ma20'] = d['close'] > d['ma20']
|
|
|
+ return d
|
|
|
+
|
|
|
+data = calc_indicators(df)
|
|
|
+
|
|
|
+# 测试的参数组合 (简化版)
|
|
|
+test_configs = [
|
|
|
+ # 配置1: 原配置
|
|
|
+ {'name': '原配置', 'adx_t': 25, 'adx_w': 30, 'ma_t': 1.002, 'ma_w': 25, 'vol_t': 0.8, 'vol_w': 20, 'volu_t': 1.5, 'volu_w': 10},
|
|
|
+ # 配置2: 提高ADX权重
|
|
|
+ {'name': '高ADX权重', 'adx_t': 25, 'adx_w': 35, 'ma_t': 1.002, 'ma_w': 20, 'vol_t': 0.8, 'vol_w': 20, 'volu_t': 1.5, 'volu_w': 10},
|
|
|
+ # 配置3: 降低ADX阈值
|
|
|
+ {'name': '低ADX阈值', 'adx_t': 20, 'adx_w': 30, 'ma_t': 1.002, 'ma_w': 25, 'vol_t': 0.8, 'vol_w': 20, 'volu_t': 1.5, 'volu_w': 10},
|
|
|
+ # 配置4: 提高MA权重
|
|
|
+ {'name': '高MA权重', 'adx_t': 25, 'adx_w': 25, 'ma_t': 1.002, 'ma_w': 30, 'vol_t': 0.8, 'vol_w': 20, 'volu_t': 1.5, 'volu_w': 10},
|
|
|
+ # 配置5: 更严格成交量
|
|
|
+ {'name': '严格成交量', 'adx_t': 25, 'adx_w': 30, 'ma_t': 1.002, 'ma_w': 25, 'vol_t': 0.8, 'vol_w': 15, 'volu_t': 2.0, 'volu_w': 15},
|
|
|
+ # 配置6: 宽松波动率
|
|
|
+ {'name': '宽松波动率', 'adx_t': 25, 'adx_w': 30, 'ma_t': 1.002, 'ma_w': 25, 'vol_t': 0.9, 'vol_w': 20, 'volu_t': 1.5, 'volu_w': 10},
|
|
|
+ # 配置7: 综合优化1
|
|
|
+ {'name': '综合优化1', 'adx_t': 22, 'adx_w': 32, 'ma_t': 1.001, 'ma_w': 28, 'vol_t': 0.85, 'vol_w': 18, 'volu_t': 1.8, 'volu_w': 12},
|
|
|
+ # 配置8: 综合优化2
|
|
|
+ {'name': '综合优化2', 'adx_t': 20, 'adx_w': 35, 'ma_t': 1.003, 'ma_w': 25, 'vol_t': 0.8, 'vol_w': 15, 'volu_t': 2.0, 'volu_w': 10},
|
|
|
+]
|
|
|
+
|
|
|
+results = []
|
|
|
+
|
|
|
+for cfg in test_configs:
|
|
|
+ print(f"\n测试: {cfg['name']}...")
|
|
|
+
|
|
|
+ scores = []
|
|
|
+ for i in range(60, len(data)):
|
|
|
+ row = data.iloc[i]
|
|
|
+
|
|
|
+ # 计算各因子得分
|
|
|
+ adx_s = cfg['adx_w'] if row['adx'] >= cfg['adx_t'] else cfg['adx_w'] * (row['adx'] / cfg['adx_t'])
|
|
|
+
|
|
|
+ if row['ma_slope'] >= cfg['ma_t'] * 1.5:
|
|
|
+ ma_s = cfg['ma_w']
|
|
|
+ elif row['ma_slope'] >= cfg['ma_t']:
|
|
|
+ ma_s = cfg['ma_w'] * 0.7
|
|
|
+ elif row['ma_slope'] >= 1.0:
|
|
|
+ ma_s = cfg['ma_w'] * 0.3
|
|
|
+ else:
|
|
|
+ ma_s = 0
|
|
|
+
|
|
|
+ if row['atr_ratio'] <= cfg['vol_t'] * 0.75:
|
|
|
+ vol_s = cfg['vol_w']
|
|
|
+ elif row['atr_ratio'] <= cfg['vol_t']:
|
|
|
+ vol_s = cfg['vol_w'] * 0.7
|
|
|
+ elif row['atr_ratio'] <= 1.0:
|
|
|
+ vol_s = cfg['vol_w'] * 0.3
|
|
|
+ else:
|
|
|
+ vol_s = 0
|
|
|
+
|
|
|
+ if row['vol_ratio'] >= cfg['volu_t'] * 1.3:
|
|
|
+ volu_s = cfg['volu_w']
|
|
|
+ elif row['vol_ratio'] >= cfg['volu_t']:
|
|
|
+ volu_s = cfg['volu_w'] * 0.7
|
|
|
+ elif row['vol_ratio'] >= 1.0:
|
|
|
+ volu_s = cfg['volu_w'] * 0.3
|
|
|
+ else:
|
|
|
+ volu_s = 0
|
|
|
+
|
|
|
+ tf_s = 15 if row['above_ma20'] else 0
|
|
|
+
|
|
|
+ total = adx_s + ma_s + vol_s + volu_s + tf_s
|
|
|
+ is_trade = total >= 60
|
|
|
+
|
|
|
+ scores.append({
|
|
|
+ 'date': data.index[i],
|
|
|
+ 'close': row['close'],
|
|
|
+ 'score': total,
|
|
|
+ 'is_trade': is_trade,
|
|
|
+ 'adx_s': adx_s, 'ma_s': ma_s, 'vol_s': vol_s, 'volu_s': volu_s, 'tf_s': tf_s
|
|
|
+ })
|
|
|
+
|
|
|
+ scores_df = pd.DataFrame(scores).set_index('date')
|
|
|
+ scores_df['ret_20d'] = scores_df['close'].pct_change(20).shift(-20) * 100
|
|
|
+
|
|
|
+ t_mask = scores_df['is_trade']
|
|
|
+ trade_days = t_mask.sum()
|
|
|
+
|
|
|
+ if trade_days > 0:
|
|
|
+ ret_20d = scores_df[t_mask]['ret_20d'].dropna()
|
|
|
+ result = {
|
|
|
+ 'name': cfg['name'],
|
|
|
+ 'adx_t': cfg['adx_t'], 'adx_w': cfg['adx_w'],
|
|
|
+ 'ma_t': cfg['ma_t'], 'ma_w': cfg['ma_w'],
|
|
|
+ 'vol_t': cfg['vol_t'], 'vol_w': cfg['vol_w'],
|
|
|
+ 'volu_t': cfg['volu_t'], 'volu_w': cfg['volu_w'],
|
|
|
+ 'trade_pct': trade_days / len(scores_df) * 100,
|
|
|
+ 'avg_score': scores_df['score'].mean(),
|
|
|
+ 'ret_20d': ret_20d.mean() if len(ret_20d) > 0 else 0,
|
|
|
+ 'win_rate': (ret_20d > 0).mean() * 100 if len(ret_20d) > 0 else 0,
|
|
|
+ 'sharpe': ret_20d.mean() / ret_20d.std() if len(ret_20d) > 0 and ret_20d.std() > 0 else 0,
|
|
|
+ }
|
|
|
+ results.append(result)
|
|
|
+ print(f" 可交易: {result['trade_pct']:.1f}%, 20日收益: {result['ret_20d']:+.2f}%, 胜率: {result['win_rate']:.1f}%")
|
|
|
+
|
|
|
+# 排序结果
|
|
|
+print("\n" + "="*70)
|
|
|
+print("参数优化结果排名")
|
|
|
+print("="*70)
|
|
|
+
|
|
|
+# 按20日收益排序
|
|
|
+results_by_return = sorted(results, key=lambda x: x['ret_20d'], reverse=True)
|
|
|
+print("\n【按20日收益排序】")
|
|
|
+print(f"{'排名':<4} {'配置名称':<12} {'可交易%':<10} {'20日收益':<12} {'胜率':<10} {'夏普':<8}")
|
|
|
+print("-"*70)
|
|
|
+for i, r in enumerate(results_by_return[:5]):
|
|
|
+ print(f"{i+1:<4} {r['name']:<12} {r['trade_pct']:>8.1f}% {r['ret_20d']:>+10.2f}% {r['win_rate']:>8.1f}% {r['sharpe']:>6.2f}")
|
|
|
+
|
|
|
+# 按胜率排序
|
|
|
+results_by_winrate = sorted(results, key=lambda x: x['win_rate'], reverse=True)
|
|
|
+print("\n【按胜率排序】")
|
|
|
+print(f"{'排名':<4} {'配置名称':<12} {'可交易%':<10} {'20日收益':<12} {'胜率':<10} {'夏普':<8}")
|
|
|
+print("-"*70)
|
|
|
+for i, r in enumerate(results_by_winrate[:5]):
|
|
|
+ print(f"{i+1:<4} {r['name']:<12} {r['trade_pct']:>8.1f}% {r['ret_20d']:>+10.2f}% {r['win_rate']:>8.1f}% {r['sharpe']:>6.2f}")
|
|
|
+
|
|
|
+# 综合最佳
|
|
|
+print("\n【综合推荐配置】")
|
|
|
+best = results_by_return[0]
|
|
|
+print(f"配置名称: {best['name']}")
|
|
|
+print(f"ADX阈值: {best['adx_t']}, 权重: {best['adx_w']}")
|
|
|
+print(f"MA斜率阈值: {best['ma_t']}, 权重: {best['ma_w']}")
|
|
|
+print(f"波动率阈值: {best['vol_t']}, 权重: {best['vol_w']}")
|
|
|
+print(f"成交量阈值: {best['volu_t']}, 权重: {best['volu_w']}")
|
|
|
+print(f"\n绩效:")
|
|
|
+print(f" 可交易比例: {best['trade_pct']:.1f}%")
|
|
|
+print(f" 20日平均收益: {best['ret_20d']:+.2f}%")
|
|
|
+print(f" 胜率: {best['win_rate']:.1f}%")
|
|
|
+print(f" 夏普比率: {best['sharpe']:.2f}")
|
|
|
+
|
|
|
+# 保存结果
|
|
|
+results_df = pd.DataFrame(results)
|
|
|
+results_df = results_df.sort_values('ret_20d', ascending=False)
|
|
|
+results_df.to_csv('/root/.openclaw/workspace/trend-quality-evaluator/param_optimization_results.csv', index=False)
|
|
|
+print("\n✓ 结果已保存: param_optimization_results.csv")
|
|
|
+
|
|
|
+print("\n" + "="*70)
|
|
|
+print("参数优化测试完成!")
|
|
|
+print("="*70)
|
openclaw