technical-patterns-lab/scripts/archive/demo_pivot_detection.py

"""
枢轴点检测可视化示例

用法：
    python scripts/demo_pivot_detection.py
"""

import numpy as np
import matplotlib.pyplot as plt
import sys
import os

# 配置中文字体
plt.rcParams['font.sans-serif'] = ['SimHei', 'Microsoft YaHei', 'Arial Unicode MS']
plt.rcParams['axes.unicode_minus'] = False

sys.path.append(os.path.join(os.path.dirname(__file__), "..", "src"))
from converging_triangle import pivots_fractal


def create_sample_data():
    """创建示例价格数据"""
    # 模拟一个对称三角形
    days = 100
    base_price = 50
    
    # 创建波动的价格数据
    high = np.zeros(days)
    low = np.zeros(days)
    
    for i in range(days):
        # 整体趋势：收敛
        upper_trend = base_price + 15 - (i / days) * 10  # 上沿下降
        lower_trend = base_price - 10 + (i / days) * 8   # 下沿上升
        
        # 添加局部波动
        wave = 3 * np.sin(i / 10) + 2 * np.sin(i / 15)
        
        high[i] = upper_trend + wave + np.random.uniform(0, 1)
        low[i] = lower_trend + wave - np.random.uniform(0, 1)
    
    return high, low


def demo_pivot_k_comparison():
    """演示不同k值对枢轴点检测的影响"""
    print("=" * 70)
    print("枢轴点检测可视化演示")
    print("=" * 70)
    
    # 生成示例数据
    high, low = create_sample_data()
    close = (high + low) / 2
    
    # 测试不同的k值
    k_values = [3, 8, 15]
    
    fig, axes = plt.subplots(len(k_values), 1, figsize=(14, 12))
    
    for idx, k in enumerate(k_values):
        ax = axes[idx]
        
        # 检测枢轴点
        ph_idx, pl_idx = pivots_fractal(high, low, k=k)
        
        # 绘制价格线
        x = np.arange(len(close))
        ax.plot(x, close, 'k-', linewidth=1, alpha=0.6, label='收盘价')
        ax.plot(x, high, 'gray', linewidth=0.5, alpha=0.3, label='最高价')
        ax.plot(x, low, 'gray', linewidth=0.5, alpha=0.3, label='最低价')
        
        # 标注枢轴点
        ax.scatter(ph_idx, high[ph_idx], 
                  marker='o', s=80, facecolors='none', 
                  edgecolors='red', linewidths=2,
                  label=f'高点枢轴 ({len(ph_idx)}个)', zorder=5)
        
        ax.scatter(pl_idx, low[pl_idx], 
                  marker='o', s=80, facecolors='none', 
                  edgecolors='green', linewidths=2,
                  label=f'低点枢轴 ({len(pl_idx)}个)', zorder=5)
        
        # 标题和标签
        ax.set_title(
            f'k = {k} (左右各{k}根K线) | '
            f'高点枢轴: {len(ph_idx)}个 | 低点枢轴: {len(pl_idx)}个',
            fontsize=12, pad=10
        )
        ax.set_ylabel('价格', fontsize=10)
        ax.legend(loc='upper right', fontsize=9)
        ax.grid(True, alpha=0.3)
        
        if idx == len(k_values) - 1:
            ax.set_xlabel('交易日', fontsize=10)
        
        print(f"\nk={k}:")
        print(f"  高点枢轴: {len(ph_idx)}个 (索引: {ph_idx[:10]}...)")
        print(f"  低点枢轴: {len(pl_idx)}个 (索引: {pl_idx[:10]}...)")
    
    plt.tight_layout()
    
    # 保存图片
    output_dir = os.path.join("docs", "images")
    os.makedirs(output_dir, exist_ok=True)
    output_path = os.path.join(output_dir, "pivot_detection_demo.png")
    plt.savefig(output_path, dpi=120)
    print(f"\n图表已保存: {output_path}")
    
    plt.show()
    
    print("\n" + "=" * 70)
    print("观察要点:")
    print("=" * 70)
    print("1. k=3: 检测到很多枢轴点，包括小幅波动")
    print("   - 优点: 捕获更多细节")
    print("   - 缺点: 容易受噪音影响")
    print()
    print("2. k=8: 中等敏感度，过滤了一些噪音")
    print("   - 平衡灵敏度和稳定性")
    print()
    print("3. k=15: 只检测显著的转折点（当前配置）")
    print("   - 优点: 稳定，适合识别主要形态")
    print("   - 缺点: 可能遗漏一些细节")
    print()
    print("推荐: 日线级别使用 k=15，分钟级别使用 k=3-5")
    print("=" * 70)


def demo_pivot_logic():
    """演示枢轴点判定逻辑"""
    print("\n" + "=" * 70)
    print("枢轴点判定逻辑示例")
    print("=" * 70)
    
    # 简单示例数据
    high = np.array([10, 12, 15, 18, 20, 17, 14, 16, 13, 11, 9])
    low = np.array([8, 9, 11, 14, 16, 13, 10, 12, 9, 7, 5])
    k = 3
    
    print(f"\n最高价: {high}")
    print(f"最低价: {low}")
    print(f"k值: {k} (左右各{k}根K线)\n")
    
    # 手动检查每个位置
    print("逐个位置检查 (索引从0开始):")
    print("-" * 70)
    
    for i in range(k, len(high) - k):
        # 检查高点
        window_high = high[i - k : i + k + 1]
        is_ph = (high[i] == np.max(window_high))
        
        # 检查低点
        window_low = low[i - k : i + k + 1]
        is_pl = (low[i] == np.min(window_low))
        
        print(f"索引 {i}:")
        print(f"  高点检查: high[{i}]={high[i]:.0f}, "
              f"窗口[{i-k}:{i+k+1}]最大值={np.max(window_high):.0f} "
              f"-> {'是枢轴高点 [YES]' if is_ph else '不是'}")
        print(f"  低点检查: low[{i}]={low[i]:.0f}, "
              f"窗口[{i-k}:{i+k+1}]最小值={np.min(window_low):.0f} "
              f"-> {'是枢轴低点 [YES]' if is_pl else '不是'}")
        print()
    
    # 使用函数检测
    ph_idx, pl_idx = pivots_fractal(high, low, k=k)
    print("-" * 70)
    print(f"函数检测结果:")
    print(f"  高点枢轴索引: {ph_idx}")
    print(f"  低点枢轴索引: {pl_idx}")
    print("=" * 70)


if __name__ == "__main__":
    # 1. 演示判定逻辑
    demo_pivot_logic()
    
    # 2. 可视化不同k值的效果
    print("\n按任意键继续查看可视化图表...")
    input()
    demo_pivot_k_comparison()