{"id":4361,"date":"2026-07-01T07:00:00","date_gmt":"2026-06-30T23:00:00","guid":{"rendered":"https:\/\/laoyulaoyu.com\/?p=4361"},"modified":"2026-06-24T14:01:46","modified_gmt":"2026-06-24T06:01:46","slug":"%e7%be%8a%e7%be%a4%e8%a1%8c%e4%b8%ba%ef%bc%88%e4%bb%8e%e4%bc%97%e5%bf%83%e7%90%86%ef%bc%89%e7%9a%84%e9%87%8f%e5%8c%96%e6%a3%80%e6%b5%8b%ef%bc%9a%e5%85%ad%e7%a7%8d%e6%96%b9%e6%b3%95%e8%af%86%e5%88%ab","status":"publish","type":"post","link":"https:\/\/laoyulaoyu.com\/index.php\/2026\/07\/01\/%e7%be%8a%e7%be%a4%e8%a1%8c%e4%b8%ba%ef%bc%88%e4%bb%8e%e4%bc%97%e5%bf%83%e7%90%86%ef%bc%89%e7%9a%84%e9%87%8f%e5%8c%96%e6%a3%80%e6%b5%8b%ef%bc%9a%e5%85%ad%e7%a7%8d%e6%96%b9%e6%b3%95%e8%af%86%e5%88%ab\/","title":{"rendered":"\u7f8a\u7fa4\u884c\u4e3a\uff08\u4ece\u4f17\u5fc3\u7406\uff09\u7684\u91cf\u5316\u68c0\u6d4b\uff1a\u516d\u79cd\u65b9\u6cd5\u8bc6\u522b\u5e02\u573a\u8fc7\u5ea6\u62e5\u6324\u4fe1\u53f7"},"content":{"rendered":"\n

\u4f5c\u8005\uff1a\u8001\u4f59\u635e\u9c7c<\/a><\/p>\n\n\n\n

\u539f\u521b\u4e0d\u6613\uff0c\u8f6c\u8f7d\u8bf7\u6807\u660e\u51fa\u5904\u53ca\u539f\u4f5c\u8005\u3002<\/mark><\/strong><\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n
\n
\ud83c\udfaf \u6838\u5fc3\u89c2\u70b9\uff1a<\/strong>\u7f8a\u7fa4\u884c\u4e3a\u662f\u73b0\u4ee3\u5e02\u573a\u4e2d\u6700\u5f3a\u5927\u7684\u529b\u91cf\u4e4b\u4e00<\/mark>\uff0c\u5b83\u4ee3\u8868\u7740\u6295\u8d44\u8005\u653e\u5f03\u79c1\u4eba\u4fe1\u606f\u8ffd\u968f\u5e02\u573a\u5171\u8bc6\u7684\u7cfb\u7edf\u6027\u504f\u5dee\u3002\u8fd9\u4e0d\u662f\u7b80\u5355\u7684\u8ddf\u98ce\uff0c\u800c\u662f\u5bfc\u81f4\u52a8\u91cf\u5f02\u5e38\u3001\u8d44\u4ea7\u6ce1\u6cab\u548c\u52a0\u901f\u5d29\u76d8\u7684\u7ed3\u6784\u6027\u5e02\u573a\u626d\u66f2\u3002<\/pre>\n<\/blockquote>\n\n\n\n
\u4e00\u3001\u4e3a\u4ec0\u4e48\u4e13\u4e1a\u6295\u8d44\u8005\u5fc5\u987b\u5173\u6ce8\u7f8a\u7fa4\u884c\u4e3a\uff1f<\/strong><\/h2>\n\n\n\n
\u4f20\u7edf\u7406\u8bba\u8ba4\u4e3a\uff0c\u6bcf\u4e2a\u6295\u8d44\u8005\u90fd\u662f\u201c\u806a\u660e\u4eba\u201d\uff0c\u5404\u81ea\u51b7\u9759\u5206\u6790\u516c\u5f00\u4fe1\u606f\uff0c\u7136\u540e\u505a\u51fa\u5bf9\u81ea\u5df1\u6700\u6709\u5229\u7684\u51b3\u7b56\uff0c\u6240\u4ee5\u5e02\u573a\u4ef7\u683c\u80fd\u53cd\u6620\u771f\u5b9e\u4ef7\u503c\u3002<\/p>\n\n\n\n
\u4f46\u73b0\u5b9e\u4e2d\uff0c\u5927\u5bb6\u5e38\u5e38\u201c\u8ddf\u98ce\u8d70\u201d\uff1a\u770b\u5230\u522b\u4eba\u4e70\u5c31\u8ddf\u7740\u4e70\uff0c\u770b\u5230\u522b\u4eba\u5356\u5c31\u8ddf\u7740\u5356\u3002\u8fd9\u79cd\u4ece\u4f17\u884c\u4e3a\u4f1a\u8ba9\u5927\u91cf\u8d44\u91d1\u76f2\u76ee\u5730\u6d8c\u5411\u6216\u9003\u79bb\u67d0\u4e2a\u8d44\u4ea7\uff0c\u7ed3\u679c\u4ef7\u683c\u6ce2\u52a8\u5b8c\u5168\u76d6\u8fc7\u4e86\u516c\u53f8\u672c\u8eab\u597d\u574f\u7b49\u57fa\u672c\u9762\u4fe1\u606f\uff0c\u5e02\u573a\u5c31\u201c\u8dd1\u504f\u201d\u4e86\u3002<\/p>\n\n\n\n
\u26a0\ufe0f \u73b0\u5b9e\u5f71\u54cd\uff1a<\/strong><\/p>\n\n\n\n
    \n
  • \u538b\u529b\u65f6\u671f30-50%\u7684\u80a1\u7968\u6536\u76ca\u65b9\u5dee\u6765\u81ea\u7f8a\u7fa4\u884c\u4e3a\u800c\u975e\u57fa\u672c\u9762\uff1b<\/li>\n\n\n\n
  • \u8d44\u672c\u8ffd\u9010\u5171\u8bc6\u53d9\u4e8b\u5bfc\u81f4\u6ce1\u6cab\u5f62\u6210\uff08\u5982AI\u3001\u52a0\u5bc6\u8d27\u5e01\u7b49\u70ed\u70b9\uff09\uff1b<\/li>\n\n\n\n
  • \u7fa4\u4f53\u53cd\u8f6c\u65f6\u56de\u64a4\u5e45\u5ea6\u653e\u59272-3\u500d\uff1b<\/li>\n\n\n\n
  • \u89e3\u91ca\u5e02\u573a\u8981\u6c42\u7684\u8d85\u989d\u6536\u76ca\u4e4b\u8c1c\u3002<\/li>\n<\/ul>\n\n\n\n
    \u4f5c\u4e3a\u91cf\u5316\u6295\u8d44\u8005\uff0c\u6211\u4eec\u9700\u8981\u7684\u4e0d\u4ec5\u662f\u5b9a\u6027\u63cf\u8ff0\uff0c\u66f4\u662f\u80fd\u591f\u91cf\u5316\u3001\u76d1\u6d4b\u548c\u9884\u8b66\u7684\u7cfb\u7edf\u5316\u5de5\u5177<\/strong>\u3002\u4e0b\u9762\u6211\u5c06\u5206\u4eab\u516d\u79cd\u7ecf\u8fc7\u5b9e\u6218\u68c0\u9a8c\u7684\u91cf\u5316\u65b9\u6cd5\u3002<\/p>\n\n\n\n
    \n
    \u5e02\u573a\u6700\u5927\u7684\u98ce\u9669\u4e0d\u662f\u6ce2\u52a8\uff0c\u800c\u662f\u5f53\u6240\u6709\u4eba\u90fd\u671d\u540c\u4e00\u4e2a\u65b9\u5411\u5954\u8dd1\u65f6\uff0c\u4f60\u4e5f\u5728\u5176\u4e2d\u3002 – \u91cf\u5316\u6295\u8d44\u7bb4\u8a00<\/p>\n<\/blockquote>\n\n\n\n
    \u4e8c\u3001\u516d\u5927\u91cf\u5316\u68c0\u6d4b\u65b9\u6cd5\u5b8c\u6574\u89e3\u6790<\/strong><\/h2>\n\n\n\n

    \u65b9\u6cd51\uff1a\u7f8a\u7fa4\u884c\u4e3a\u6307\u6570\uff08HBI\uff09 – \u504f\u79bb\u5ea6\u68c0\u6d4b\u5668<\/strong><\/h3>\n\n\n\n

    \u6838\u5fc3\u539f\u7406\u4e0e\u6570\u5b66\u516c\u5f0f<\/strong><\/h4>\n\n\n\n
    HBI\u91cf\u5316\u4e2a\u80a1\u4e0e\u57fa\u51c6\u4e4b\u95f4\u7684\u76f8\u5bf9\u6536\u76ca\u79bb\u6563\u5ea6<\/strong>\uff0c\u516c\u5f0f\u5982\u4e0b\uff1a<\/p>\n\n\n\n
    HBI_i,t(w) = |\u5e73\u5747\u6536\u76ca\u7387_i,t(w) – \u5e73\u5747\u6536\u76ca\u7387_m,t(w)| \/ |\u5e73\u5747\u6536\u76ca\u7387_m,t(w)|<\/strong><\/p>\n\n\n\n
    \u5176\u4e2d\uff1a<\/p>\n\n\n\n
      \n
    • \u5e73\u5747\u6536\u76ca\u7387_i,t(w)\u00a0= <\/strong>\u8bc1\u5238i\u5728\u7a97\u53e3w\u5185\u7684\u5e73\u5747\u6536\u76ca\u7387<\/li>\n\n\n\n
    • \u5e73\u5747\u6536\u76ca\u7387_m,t(w)\u00a0= <\/strong>\u57fa\u51c6\u5728\u7a97\u53e3w\u5185\u7684\u5e73\u5747\u6536\u76ca\u7387<\/li>\n<\/ul>\n\n\n\n
      \u4e09\u79cd\u5178\u578b\u5e02\u573a\u72b6\u6001<\/strong><\/h4>\n\n\n\n
      HBI\u8303\u56f4<\/th>\u5e02\u573a\u72b6\u6001<\/th>\u6295\u8d44\u542b\u4e49<\/th>\u64cd\u4f5c\u5efa\u8bae<\/th><\/tr>
      HBI < 0.3<\/td>\u6781\u7aef\u4e00\u81f4\u6027<\/td>\u7fa4\u4f53\u9677\u9631<\/td>\u5efa\u8bae\u89c4\u907f<\/td><\/tr>
      0.3 \u2264 HBI \u2264 2.0<\/td>\u6b63\u5e38\u8ddf\u8e2a<\/td>\u4e2d\u6027\u914d\u7f6e<\/td>\u7ef4\u6301\u73b0\u72b6<\/td><\/tr>
      HBI > 2.0<\/td>\u7edf\u8ba1\u5f02\u5e38\u503c<\/td>\u72ec\u7acb\u6295\u8d44\u673a\u4f1a<\/td>\u53ef\u8003\u8651\u589e\u52a0\u914d\u7f6e<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
      \"\"<\/figure>\n\n\n\n
      HBI\u6307\u6807\u591a\u65f6\u95f4\u7a97\u53e3\u53ef\u89c6\u5316 – \u663e\u793a\u82f1\u4f1f\u8fbe\u4e0e\u7eb3\u65af\u8fbe\u514b100\u7684\u76f8\u5bf9\u504f\u79bb\u5ea6<\/p>\n\n\n\n
      Python\u5b9e\u73b0\u4ee3\u7801<\/strong><\/h3>\n\n\n\n
      import yfinance as yf<\/code>import pandas as pd<\/code>import numpy as np<\/code># ------------------------------<\/code># 1. \u4e0b\u8f7d\u6570\u636e<\/code># ------------------------------<\/code>nvda = yf.download(\"NVDA\", start=\"2018-01-01\", end=\"2026-03-14\")[\"Close\"]<\/code>qqq  = yf.download(\"QQQ\",  start=\"2018-01-01\", end=\"2026-03-14\")[\"Close\"]<\/code># ------------------------------<\/code># 2. \u8ba1\u7b97\u5bf9\u6570\u6536\u76ca\u7387\uff08\u589e\u5f3a\u5e73\u7a33\u6027\uff09<\/code># ------------------------------<\/code>nvda_r = np.log(nvda \/ nvda.shift(1)).dropna()<\/code>qqq_r  = np.log(qqq  \/ qqq.shift(1)).dropna()<\/code># ------------------------------<\/code># 3. \u591a\u65f6\u95f4\u7a97\u53e3 HBI\uff08\u7f8a\u7fa4\u884c\u4e3a\u6307\u6570\uff09\u8ba1\u7b97<\/code># ------------------------------<\/code>windows = [20, 60, 120]          # \u77ed\u671f \/ \u4e2d\u671f \/ \u957f\u671f<\/code>hbi_results = {}<\/code>for w in windows:<\/code>    # \u5206\u5b50\uff1aNVDA \u4e0e QQQ \u6536\u76ca\u7387\u5dee\u5f02\u7684\u7edd\u5bf9\u503c\u7684\u6eda\u52a8\u5747\u503c<\/code>    # \u5206\u6bcd\uff1aQQQ \u6536\u76ca\u7387\u7edd\u5bf9\u503c\u7684\u6eda\u52a8\u5747\u503c<\/code>    hbi = (np.abs(nvda_r - qqq_r).rolling(w).mean()<\/code>           \/ np.abs(qqq_r).rolling(w).mean())<\/code>    hbi_results[f\"HBI_{w}d\"] = hbi<\/code># \u6b64\u65f6 hbi_results \u662f\u4e00\u4e2a\u5b57\u5178\uff0c\u952e\u4e3a 'HBI_20d'\u3001'HBI_60d'\u3001'HBI_120d'<\/code># \u6bcf\u4e2a\u952e\u5bf9\u5e94\u7684\u503c\u4e3a\u4e00\u4e2a pandas Series\uff0c\u7d22\u5f15\u4e3a\u65e5\u671f\uff0c\u503c\u4e3a\u8be5\u7a97\u53e3\u4e0b\u7684 HBI \u5e8f\u5217<\/code><\/code><\/pre>\n\n\n\n
      \ud83d\udcca \u5b9e\u6218\u4fe1\u53f7\u9a8c\u8bc1\uff1a<\/strong><\/p>\n\n\n\n
        \n
      • 2020\u5e743\u6708\uff08\u75ab\u60c5\u5e95\u90e8\uff09<\/strong>\uff1aHBI_60d = 1.8 \u2192 \u65e9\u671f\u590d\u82cf\u4fe1\u53f7\uff1b<\/li>\n\n\n\n
      • 2022\u5e741\u6708\uff08\u79d1\u6280\u80a1\u5cf0\u503c\uff09<\/strong>\uff1aHBI_60d \u2192 0.25 \u2192 \u7fa4\u4f53\u4e00\u81f4\u6027\u9884\u8b66\uff1b<\/li>\n\n\n\n
      • 2023\u5e746\u6708\uff08AI\u7a81\u7834\uff09<\/strong>\uff1aHBI_120d = 3.7 \u2192 800%\u8d85\u989d\u6536\u76ca\u786e\u8ba4\uff1b<\/li>\n\n\n\n
      • 2025\u5e7412\u6708<\/strong>\uff1aHBI_20d = 2.9 \u2192 \u5f53\u524d\u504f\u79bb\u4fe1\u53f7\u3002<\/li>\n<\/ul>\n\n\n\n
        \u65b9\u6cd52\uff1a\u6a2a\u622a\u9762\u7edd\u5bf9\u504f\u5dee\uff08CSAD\uff09 – \u6295\u8d44\u7ec4\u5408\u96c6\u4e2d\u5ea6\u96f7\u8fbe<\/strong><\/h3>\n\n\n\n

        \u7406\u8bba\u57fa\u7840\u4e0e\u516c\u5f0f<\/strong><\/h4>\n\n\n\n
        CSAD\u8861\u91cf\u6295\u8d44\u7ec4\u5408\u4e2d\u4e2a\u80a1\u6536\u76ca\u76f8\u5bf9\u4e8e\u5e02\u573a\u5e73\u5747\u6536\u76ca\u7684\u79bb\u6563\u7a0b\u5ea6<\/strong>\uff1a<\/p>\n\n\n\n
        CSAD_t(w) = [N\/(N-1)] \u00d7 \u03a3|R_i,t(w) – \u5e73\u5747\u6536\u76ca\u7387_m,t(w)|<\/p>\n\n\n\n
        \u76f4\u89c2\u89e3\u91ca\u4e0e\u6295\u8d44\u542b\u4e49<\/strong><\/h4>\n\n\n\n
          \n
        • \u4f4eCSAD\u00a0= <\/strong>\u6536\u76ca\u6536\u655b \u2192\u00a0\u7f8a\u7fa4\u884c\u4e3a\uff08\u98ce\u9669\u533a\u57df\uff09<\/li>\n\n\n\n
        • \u9ad8CSAD\u00a0= <\/strong>\u6536\u76ca\u79bb\u6563 \u2192\u00a0\u72ec\u7acb\u51b3\u7b56\uff08\u673a\u4f1a\u533a\u57df\uff09<\/li>\n<\/ul>\n\n\n\n
          \u5176\u4e2dN\/(N-1)\u9879\u7528\u4e8e\u6821\u6b63\u6709\u9650\u6837\u672c\u504f\u5dee\u3002<\/p>\n\n\n\n
          \"\"<\/figure>\n\n\n\n
          CSAD\u6307\u6807\u5386\u53f2\u8d70\u52bf – \u663e\u793a\u4e03\u5927\u79d1\u6280\u80a1\u76f8\u5bf9\u4e8e\u6807\u666e500\u7684\u79bb\u6563\u5ea6<\/p>\n\n\n\n
          \u4e03\u5927\u79d1\u6280\u80a1\u5b9e\u73b0\u4ee3\u7801<\/strong><\/h4>\n\n\n\n
          import yfinance as yf<\/code>import pandas as pd<\/code>import numpy as np<\/code><\/code># ------------------------------<\/code># 1. \u5b9a\u4e49\u4e03\u5927\u79d1\u6280\u80a1<\/code># ------------------------------<\/code>mag7 = [\"NVDA\", \"AAPL\", \"MSFT\", \"GOOGL\", \"META\", \"AMD\", \"TSLA\"]<\/code><\/code># ------------------------------<\/code># 2. \u4e0b\u8f7d\u6807\u666e500\u6307\u6570\uff08\u65e5\u6536\u76ca\u7387\uff09<\/code># ------------------------------<\/code>spx = yf.download(\"^GSPC\", start=\"2018-01-01\")[\"Close\"].pct_change()<\/code><\/code># ------------------------------<\/code># 3. \u751f\u4ea7\u7ea7 CSAD\uff08\u6a2a\u622a\u9762\u7edd\u5bf9\u504f\u5dee\uff09\u8ba1\u7b97\u5668<\/code>#    \u7528\u4e8e\u8861\u91cf\u4e2a\u80a1\u6536\u76ca\u7387\u4e0e\u5e02\u573a\u5e73\u5747\u6536\u76ca\u7387\u7684\u504f\u79bb\u7a0b\u5ea6<\/code># ------------------------------<\/code>def compute_csad(stocks_df, market_sr, window=20):<\/code>    \"\"\"<\/code>    \u8ba1\u7b97\u622a\u9762\u7edd\u5bf9\u504f\u5dee\uff08Cross-Sectional Absolute Deviation\uff09<\/code><\/code>    \u53c2\u6570:<\/code>        stocks_df : DataFrame\uff0c\u6bcf\u5217\u4e3a\u4e00\u53ea\u80a1\u7968\u7684\u6536\u76ca\u7387\u5e8f\u5217\uff08\u7d22\u5f15\u4e3a\u65e5\u671f\uff09<\/code>        market_sr : Series\uff0c\u5e02\u573a\u57fa\u51c6\u6536\u76ca\u7387\u5e8f\u5217\uff08\u7d22\u5f15\u4e3a\u65e5\u671f\uff09<\/code>        window    : int\uff0c\u6eda\u52a8\u7a97\u53e3\u5927\u5c0f<\/code><\/code>    \u8fd4\u56de:<\/code>        pd.Series\uff0c\u7d22\u5f15\u4e3a\u6bcf\u4e2a\u7a97\u53e3\u7684\u7ed3\u675f\u65e5\u671f\uff0c\u503c\u4e3a\u5bf9\u5e94\u7684 CSAD<\/code>    \"\"\"<\/code><\/code>    # \u786e\u4fdd\u4e24\u4e2a\u6570\u636e\u7684\u7d22\u5f15\u5bf9\u9f50\uff08\u53ea\u4fdd\u7559\u5171\u540c\u65e5\u671f\uff09<\/code>    common_idx = stocks_df.index.intersection(market_sr.index)<\/code>    stocks_df = stocks_df.loc[common_idx]<\/code>    market_sr = market_sr.loc[common_idx]<\/code><\/code>    csad_values = []<\/code>    # \u4ece window \u5f00\u59cb\u6eda\u52a8<\/code>    for i in range(window, len(stocks_df)):<\/code>        # \u53d6\u51fa\u5f53\u524d\u7a97\u53e3\u5185\u7684\u80a1\u7968\u6536\u76ca\u7387<\/code>        window_stocks = stocks_df.iloc[i - window : i]<\/code>        # \u5f53\u524d\u7a97\u53e3\u5185\u5e02\u573a\u6536\u76ca\u7387\u7684\u5747\u503c<\/code>        window_mkt_mean = market_sr.iloc[i - window : i].mean()<\/code>        # \u4e2a\u80a1\u6570\u91cf<\/code>        N = len(stocks_df.columns)<\/code>        # \u8ba1\u7b97 CSAD = (N\/(N-1)) * \u5e73\u5747\u7edd\u5bf9\u504f\u5dee<\/code>        # \u5bf9\u6bcf\u53ea\u80a1\u7968\u6c42 |\u4e2a\u80a1\u6536\u76ca\u7387 - \u5e02\u573a\u5747\u503c|\uff0c\u518d\u5728\u6240\u6709\u80a1\u7968\u548c\u6240\u6709\u65e5\u671f\u4e0a\u53d6\u5e73\u5747<\/code>        csad = (N \/ (N - 1)) * np.abs(window_stocks.sub(window_mkt_mean, axis=0)).mean().mean()<\/code>        csad_values.append(csad)<\/code><\/code>    # \u8fd4\u56de Series\uff0c\u7d22\u5f15\u4e3a\u539f DataFrame \u4ece window \u4f4d\u7f6e\u5f00\u59cb\u7684\u65e5\u671f<\/code>    return pd.Series(csad_values, index=stocks_df.index[window:])<\/code><\/code><\/pre>\n\n\n\n
          \ud83d\udcc8 \u5386\u53f2\u9a8c\u8bc1\u4e0e\u5173\u952e\u4fe1\u53f7\uff1a<\/strong><\/p>\n\n\n\n
            \n
          • 2022\u5e741\u6708<\/strong>\uff1aCSAD_20d = \u7b2c4\u767e\u5206\u4f4d \u2192 \u79d1\u6280\u80a1\u5cf0\u503c\u786e\u8ba4\uff1b<\/li>\n\n\n\n
          • 2020\u5e743\u6708<\/strong>\uff1aCSAD_20d\u6fc0\u589e \u2192 \u75ab\u60c5\u8f6e\u52a8\u673a\u4f1a\uff1b<\/li>\n\n\n\n
          • 2023\u5e7411\u6708<\/strong>\uff1aCSAD_120d\u4f4e\u8c37 \u2192 AI\u9886\u5bfc\u5730\u4f4d\u5de9\u56fa\u3002<\/li>\n<\/ul>\n\n\n\n
            \u91cd\u8981\u6295\u8d44\u539f\u5219<\/strong>\uff1a\u65e0\u8bba\u52a8\u91cf\u5982\u4f55\uff0cCSAD < \u7b2c5\u767e\u5206\u4f4d = \u7cfb\u7edf\u6027\u964d\u4f4e\u98ce\u9669\u655e\u53e3<\/strong>\u3002<\/p>\n\n\n\n
            \u65b9\u6cd53\uff1a\u5206\u4f4d\u6570\u56de\u5f52\u5206\u6790 – \u6781\u7aef\u60c5\u51b5\u884c\u4e3a\u68c0\u6d4b<\/strong><\/h3>\n\n\n\n

            \u8d85\u8d8a\u4f20\u7edf\u5747\u503c\u56de\u5f52<\/strong><\/h4>\n\n\n\n
            \u4f20\u7edfOLS\u56de\u5f52\u6355\u6349\u5e73\u5747\u6761\u4ef6\uff0c\u800c\u5206\u4f4d\u6570\u56de\u5f52\u63ed\u793a\u6781\u7aef\u60c5\u51b5<\/strong>\uff0c\u5176\u4e2d\u7f8a\u7fa4\u884c\u4e3a\u88ab\u653e\u5927\uff1a<\/p>\n\n\n\n
            \u5206\u4f4d\u6570<\/th>\u56de\u5f52\u7cfb\u6570<\/th>\u5e02\u573a\u542b\u4e49<\/th>\u98ce\u9669\u63d0\u793a<\/th><\/tr>
            Q05 (5%)<\/td>\u03b2 < -1.5<\/td>\u5d29\u76d8\u653e\u5927\u5668<\/td>\u635f\u5931\u8d85\u8fc7\u57fa\u51c61.5\u500d\u4ee5\u4e0a<\/td><\/tr>
            Q95 (95%)<\/td>\u03b2 > 1.5<\/td>\u4e0a\u6da8\u53c2\u4e0e\u8005<\/td>\u9886\u5bfc\u5730\u4f4d\u786e\u8ba4<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
            \"\"<\/figure>\n\n\n\n
            \u82f1\u4f1f\u8fbe\u5206\u4f4d\u6570\u56de\u5f52\u5206\u6790 – \u663e\u793a\u4e0d\u540c\u5e02\u573a\u6761\u4ef6\u4e0b\u7684\u4e0d\u5bf9\u79f0\u884c\u4e3a<\/p>\n\n\n\n
            \u5b8c\u6574\u5206\u4f4d\u6570\u5206\u6790\u6846\u67b6<\/strong><\/h4>\n\n\n\n
              \n
            • Q05\uff1a\u6781\u7aef\u4e0b\u884c\uff08\u5d29\u76d8\u56de\u5f52\u7cfb\u6570\uff09<\/strong><\/li>\n\n\n\n
            • Q25\uff1a\u4e0b\u56db\u5206\u4f4d<\/strong><\/li>\n\n\n\n
            • Q50\uff1a\u4e2d\u4f4d\u6570\uff08\u4f20\u7edf\u56de\u5f52\u7cfb\u6570\u7b49\u6548\uff09<\/strong><\/li>\n\n\n\n
            • Q75\uff1a\u4e0a\u56db\u5206\u4f4d<\/strong><\/li>\n\n\n\n
            • Q95\uff1a\u6781\u7aef\u4e0a\u884c\uff08\u4e0a\u6da8\u56de\u5f52\u7cfb\u6570\uff09<\/strong><\/li>\n<\/ul>\n\n\n\n
              \u82f1\u4f1f\u8fbe\u5c3e\u90e8\u5206\u6790\u5b9e\u73b0<\/strong><\/h4>\n\n\n\n
              import statsmodels.formula.api as smf<\/code><\/code># ------------------------------<\/code># 1. \u51c6\u5907\u5206\u4f4d\u6570\u56de\u5f52\u6570\u636e<\/code>#    \u4ece returns \u4e2d\u63d0\u53d6 NVDA \u548c\u6807\u666e500 \u65e5\u6536\u76ca\u7387\uff0c\u5e76\u5220\u9664\u7f3a\u5931\u503c<\/code># ------------------------------<\/code>qr_data = returns[[\"NVDA\", \"^GSPC\"]].dropna()<\/code>qr_df = qr_data.copy()<\/code>qr_df.columns = [\"NVDA_ret\", \"SPX_ret\"]   # \u91cd\u547d\u540d\u5217\uff0c\u4fbf\u4e8e\u516c\u5f0f\u5f15\u7528<\/code><\/code># ------------------------------<\/code># 2. \u8bbe\u5b9a\u5f85\u4f30\u8ba1\u7684\u5206\u4f4d\u6570\u548c\u6eda\u52a8\u7a97\u53e3\u957f\u5ea6\uff08126\u4e2a\u4ea4\u6613\u65e5 \u2248 6\u4e2a\u6708\uff09<\/code># ------------------------------<\/code>quantiles = [0.05, 0.25, 0.50, 0.75, 0.95]<\/code>window = 126<\/code><\/code># ------------------------------<\/code># 3. \u5bf9\u6bcf\u4e2a\u5206\u4f4d\u6570\u8fdb\u884c\u5206\u4f4d\u6570\u56de\u5f52\uff0c\u8f93\u51fa\u6700\u8fd1 window \u4e2a\u6837\u672c\u7684\u7cfb\u6570<\/code>#    \u6ce8\u610f\uff1a\u4f7f\u7528 iloc \u660e\u786e\u6309\u4f4d\u7f6e\u53d6\u6700\u540e window \u884c\uff0c\u907f\u514d\u5207\u7247\u6b67\u4e49<\/code># ------------------------------<\/code>for q in quantiles:<\/code>    # \u53d6\u6700\u8fd1 window \u4e2a\u4ea4\u6613\u65e5\u7684\u6570\u636e<\/code>    sample = qr_df.iloc[-window:]    # \u66f4\u6e05\u6670\u7684\u5199\u6cd5<\/code>    # \u62df\u5408\u5206\u4f4d\u6570\u56de\u5f52<\/code>    model = smf.quantreg(\"NVDA_ret ~ SPX_ret\", sample).fit(q=q)<\/code>    # \u8f93\u51fa\u5206\u4f4d\u70b9\u5bf9\u5e94\u7684\u56de\u5f52\u7cfb\u6570\uff08\u659c\u7387\uff09<\/code>    print(f\"Q{q:.0%} \u56de\u5f52\u7cfb\u6570: {model.params['SPX_ret']:.3f}\")<\/code><\/code><\/pre>\n\n\n\n
              \u26a0\ufe0f \u5173\u952e2022\u5e74\u98ce\u9669\u4fe1\u53f7\uff1a<\/strong><\/p>\n\n\n\n
              Q05\u56de\u5f52\u7cfb\u6570 = -2.8 \u2192 \u82f1\u4f1f\u8fbe\u5728\u5d29\u76d8\u671f\u95f4\u635f\u5931\u8d85\u8fc7\u6807\u666e500\u6307\u65702.8\u500d<\/p>\n\n\n\n
              \u65b9\u6cd54\uff1a\u52a8\u6001\u76f8\u5173\u6027\u7f51\u7edc – \u53ef\u89c6\u5316\u884c\u4e3a\u805a\u7c7b<\/strong><\/h3>\n\n\n\n

              \u4ece\u77e9\u9635\u5230\u884c\u4e3a\u805a\u7c7b<\/strong><\/h4>\n\n\n\n
              \u7f51\u7edc\u65b9\u6cd5\u8bba\uff1a<\/p>\n\n\n\n
                \n
              • \u8282\u70b9\uff1a<\/strong>\u5355\u4e2a\u8bc1\u5238<\/li>\n\n\n\n
              • \u8fb9\uff1a<\/strong>|\u76f8\u5173\u6027| > 0.6\uff08\u4fdd\u5b88\u9608\u503c\uff09<\/strong><\/li>\n\n\n\n
              • \u5bc6\u5ea6\uff1a<\/strong>\u8fb9\u6570 \/ [\u8282\u70b9\u6570 \u00d7 (\u8282\u70b9\u6570-1)\/2]<\/li>\n\n\n\n
              • \u4e2d\u5fc3\u6027\uff1a<\/strong>\u6bcf\u53ea\u80a1\u7968\u7684\u8fde\u63a5\u6027\u5ea6\u91cf<\/li>\n<\/ul>\n\n\n\n
                \"\"<\/figure>\n\n\n\n
                \u79d1\u6280\u80a1\u52a8\u6001\u76f8\u5173\u6027\u7f51\u7edc – \u53ef\u89c6\u5316\u5e02\u573a\u884c\u4e3a\u805a\u7c7b\u7a0b\u5ea6<\/p>\n\n\n\n
                Python\u7f51\u7edc\u5206\u6790\u5b9e\u73b0<\/strong><\/h3>\n\n\n\n
                import networkx as nx<\/code>from itertools import combinations   # \u63a8\u8350\u4f7f\u7528\uff0c\u907f\u514d\u91cd\u590d\u904d\u5386<\/code><\/code># ------------------------------------------------<\/code># 1. \u8ba1\u7b97\u6700\u8fd1 90 \u4e2a\u4ea4\u6613\u65e5\u7684\u52a8\u6001\u76f8\u5173\u7cfb\u6570\u77e9\u9635<\/code>#    rolling(90).corr() \u8fd4\u56de MultiIndex DataFrame\uff0c<\/code>#    \u7528 .iloc[-1] \u53d6\u51fa\u6700\u540e\u4e00\u4e2a\u65f6\u95f4\u622a\u9762\uff08\u5373\u6700\u65b0\u65e5\u671f\u4e0b\u7684\u76f8\u5173\u7cfb\u6570\u77e9\u9635\uff09<\/code># ------------------------------------------------<\/code>corr_matrix = returns[tech_universe].rolling(90).corr().iloc[-1]<\/code><\/code># ------------------------------------------------<\/code># 2. \u6784\u5efa\u65e0\u5411\u7f51\u7edc\uff08\u9608\u503c\u8fc7\u6ee4\uff1a|\u76f8\u5173\u7cfb\u6570| > 0.6\uff09<\/code>#    \u53ea\u904d\u5386\u4e0d\u91cd\u590d\u7684\u80a1\u7968\u7ec4\u5408 (t1, t2)\uff0c\u907f\u514d i->j \u548c j->i \u91cd\u590d\u6dfb\u52a0<\/code># ------------------------------------------------<\/code>G = nx.Graph()<\/code><\/code>for t1, t2 in combinations(tech_universe, 2):<\/code>    corr_val = corr_matrix.loc[t1, t2]<\/code>    if abs(corr_val) > 0.6:          # \u5f3a\u6b63\u76f8\u5173\u6216\u5f3a\u8d1f\u76f8\u5173\u5747\u89c6\u4e3a\u6709\u6548\u8fde\u63a5<\/code>        G.add_edge(t1, t2, weight=corr_val)<\/code># ------------------------------------------------<\/code># 3. \u8ba1\u7b97\u7f51\u7edc\u5bc6\u5ea6\u5e76\u8f93\u51fa<\/code>#    \u5bc6\u5ea6 = \u5b9e\u9645\u8fb9\u6570 \/ \u53ef\u80fd\u7684\u6700\u5927\u8fb9\u6570\uff0c\u53cd\u6620\u6574\u4f53\u8fde\u63a5\u7d27\u5bc6\u5ea6<\/code># ------------------------------------------------<\/code>density = nx.density(G)<\/code>print(f\"\u7f51\u7edc\u5bc6\u5ea6: {density:.1%}\")<\/code><\/code><\/pre>\n\n\n\n
                \u5bc6\u5ea6\u72b6\u6001\u6846\u67b6\u4e0e\u6295\u8d44\u5efa\u8bae<\/strong><\/h4>\n\n\n\n
                \u5bc6\u5ea6\u8303\u56f4<\/th>\u5e02\u573a\u72b6\u6001<\/th>\u6295\u8d44\u5efa\u8bae<\/th>\u98ce\u9669\u7b49\u7ea7<\/th><\/tr>
                \u5bc6\u5ea6 < 40%<\/td>\u5206\u6563\u5b9a\u4f4d<\/td>\u53ef\u5b89\u5168\u589e\u52a0\u914d\u7f6e<\/td>\u4f4e\u98ce\u9669<\/td><\/tr>
                40-70%<\/td>\u9002\u5ea6\u805a\u7c7b<\/td>\u5bc6\u5207\u76d1\u63a7<\/td>\u4e2d\u7b49\u98ce\u9669<\/td><\/tr>
                > 70%<\/td>\u5355\u4e00\u8d85\u5927\u4ea4\u6613<\/td>\u9000\u51fa\u4fe1\u53f7<\/td>\u9ad8\u98ce\u9669<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
                \ud83d\udcc8 2022\u5e74\u79d1\u6280\u80a1\u5cf0\u503c\u89c2\u5bdf\uff1a<\/strong><\/p>\n\n\n\n
                \u7f51\u7edc\u5bc6\u5ea6 = 82% \u2192 \u79d1\u6280\u80a1\u6210\u4e3a”\u5355\u4e00\u4ea4\u6613”\uff0c\u5e02\u573a\u8fc7\u5ea6\u96c6\u4e2d<\/p>\n\n\n\n
                \u65b9\u6cd55\uff1a\u9690\u9a6c\u5c14\u53ef\u592b\u6a21\u578b – \u4e0d\u53ef\u89c2\u6d4b\u72b6\u6001\u8bc6\u522b<\/strong><\/h3>\n\n\n\n

                \u53d1\u73b0\u4e0d\u53ef\u89c2\u6d4b\u5e02\u573a\u72b6\u6001<\/strong><\/h4>\n\n\n\n
                HMM\u8f93\u5165\u7279\u5f81\uff1a[HBI_60d, CSAD_20d, CSAD_120d, \u7f51\u7edc\u5bc6\u5ea6]<\/strong><\/p>\n\n\n\n
                \"\"<\/figure>\n\n\n\n
                HMM\u56db\u72b6\u6001\u5206\u7c7b – \u663e\u793a\u4e0d\u540c\u7f8a\u7fa4\u884c\u4e3a\u7a0b\u5ea6\u7684\u5e02\u573a\u72b6\u6001<\/p>\n\n\n\n
                \u56db\u72b6\u6001\u5206\u7c7b\u7cfb\u7edf<\/strong><\/h3>\n\n\n\n
                  \n
                • \u72b6\u60010\uff08\u7eff\u8272\uff09<\/strong>\uff1a\u4f4e\u7f8a\u7fa4\u884c\u4e3a – \u5206\u6563\u5b9a\u4f4d<\/li>\n\n\n\n
                • \u72b6\u60011\uff08\u9ec4\u8272\uff09<\/strong>\uff1a\u9002\u5ea6\u4e00\u81f4\u6027 – \u4e2d\u6027<\/li>\n\n\n\n
                • \u72b6\u60012\uff08\u6a59\u8272\uff09<\/strong>\uff1a\u9ad8\u5ea6\u96c6\u4e2d – \u8c28\u614e<\/li>\n\n\n\n
                • \u72b6\u60013\uff08\u7ea2\u8272\uff09<\/strong>\uff1a\u6781\u7aef\u7f8a\u7fa4\u884c\u4e3a – \u53cd\u8f6c\u5373\u5c06\u53d1\u751f<\/li>\n<\/ul>\n\n\n\n
                  Python\u5b9e\u73b0<\/strong><\/h3>\n\n\n\n
                  from hmmlearn.hmm import GaussianHMM<\/code>from sklearn.preprocessing import StandardScaler<\/code><\/code># ------------------------------------------------<\/code># 1. \u6784\u5efa\u7279\u5f81\u6570\u636e\u96c6\uff08\u786e\u4fdd\u5404\u5e8f\u5217\u7d22\u5f15\u4e00\u81f4\uff0c\u7f3a\u5931\u503c\u81ea\u52a8\u5bf9\u9f50\uff09<\/code>#    \u5305\u542b\uff1a60\u5929HBI\u7f8a\u7fa4\u884c\u4e3a\u6307\u6570\u3001\u77ed\u671fCSAD\u3001\u957f\u671fCSAD\u3001\u7f51\u7edc\u5bc6\u5ea6<\/code># ------------------------------------------------<\/code>features = pd.DataFrame({<\/code>    \"hbi\": hbi_dict[\"HBI_60d\"],      # \u6765\u81ea\u4e4b\u524d\u8ba1\u7b97\u7684HBI\u5b57\u5178<\/code>    \"csad_short\": csad_short,        # \u77ed\u671f\uff08\u598220\u65e5\uff09CSAD\u5e8f\u5217<\/code>    \"csad_long\": csad_long,          # \u957f\u671f\uff08\u5982120\u65e5\uff09CSAD\u5e8f\u5217<\/code>    \"density\": density_series        # \u7f51\u7edc\u5bc6\u5ea6\u65f6\u95f4\u5e8f\u5217<\/code>}).dropna()   # \u5220\u9664\u4efb\u4e00\u6307\u6807\u7f3a\u5931\u7684\u65f6\u70b9\uff0c\u4fdd\u8bc1\u6a21\u578b\u8f93\u5165\u5b8c\u6574<\/code><\/code># ------------------------------------------------<\/code># 2. \u6807\u51c6\u5316\u7279\u5f81\uff08\u4f7f\u5404\u6307\u6807\u5747\u503c\u4e3a0\u3001\u65b9\u5dee\u4e3a1\uff09<\/code>#    \u907f\u514d\u91cf\u7eb2\u5dee\u5f02\u5f71\u54cdHMM\u7684\u534f\u65b9\u5dee\u4f30\u8ba1<\/code># ------------------------------------------------<\/code>X = StandardScaler().fit_transform(features)<\/code><\/code># ------------------------------------------------<\/code># 3. \u6784\u5efa\u5e76\u8bad\u7ec3\u9ad8\u65af\u9690\u9a6c\u5c14\u53ef\u592b\u6a21\u578b<\/code>#    \u8bbe\u7f6e4\u4e2a\u9690\u85cf\u72b6\u6001\uff0c\u8fed\u4ee31000\u6b21\uff0c\u56fa\u5b9a\u968f\u673a\u79cd\u5b50\u4fdd\u8bc1\u53ef\u590d\u73b0<\/code># ------------------------------------------------<\/code>hmm = GaussianHMM(n_components=4, n_iter=1000, random_state=42)<\/code>regimes = hmm.fit_predict(X)   # \u62df\u5408\u6570\u636e\u5e76\u9884\u6d4b\u6bcf\u4e2a\u6837\u672c\u70b9\u7684\u72b6\u6001\uff080~3\uff09<\/code><\/code># \u8f93\u51fa\u72b6\u6001\u6807\u7b7e\u7684\u957f\u5ea6\uff0c\u4ee5\u4fbf\u68c0\u67e5<\/code>print(f\"\u9690\u72b6\u6001\u5e8f\u5217\u957f\u5ea6: {len(regimes)}\uff0c\u4e0e\u7279\u5f81\u884c\u6570\u4e00\u81f4: {len(features)}\")<\/code><\/code><\/pre>\n\n\n\n
                  \ud83d\udd04 \u72b6\u6001\u8f6c\u6362\u6982\u7387\u6d1e\u5bdf\uff1a<\/strong><\/p>\n\n\n\n
                  \u6781\u7aef \u2192 \u6781\u7aef\uff1a92%<\/strong>\uff08\u4e00\u65e6\u96c6\u4e2d\uff0c\u4fdd\u6301\u96c6\u4e2d\uff09\u3002<\/p>\n\n\n\n
                  \u4f4e \u2192 \u9002\u5ea6\uff1a67%<\/strong>\uff08\u9010\u6b65\u5efa\u7acb\u4e00\u81f4\u6027\uff09\u3002<\/p>\n\n\n\n
                  \u65b9\u6cd56\uff1a\u7efc\u5408\u4fe1\u53f7\u6846\u67b6 – \u591a\u7ef4\u4fe1\u606f\u6574\u5408<\/strong><\/h3>\n\n\n\n

                  \u590d\u5408\u4fe1\u53f7\u6784\u5efa\u516c\u5f0f<\/strong><\/h4>\n\n\n\n
                  \u7efc\u5408\u4fe1\u53f7_t = 0.30\u00d7HBI_\u504f\u79bb + 0.25\u00d7CSAD_\u6781\u7aef + 0.25\u00d7Q05_\u56de\u5f52\u7cfb\u6570_\u5b89\u5168 + 0.20\u00d7\u5bc6\u5ea6_\u4f4e<\/p>\n\n\n\n
                  \"\"<\/figure>\n\n\n\n
                  \u7efc\u5408\u4fe1\u53f7\u751f\u6210\u6846\u67b6 – \u6574\u5408\u591a\u7ef4\u5ea6\u4fe1\u606f\u751f\u6210\u4ea4\u6613\u4fe1\u53f7<\/p>\n\n\n\n
                  \u914d\u7f6e\u89c4\u6a21\u51b3\u7b56\u89c4\u5219<\/strong><\/h3>\n\n\n\n
                  \u4fe1\u53f7\u6761\u4ef6<\/th>\u64cd\u4f5c<\/th>\u914d\u7f6e\u89c4\u6a21<\/th>\u98ce\u9669\u7b49\u7ea7<\/th><\/tr>
                  \u4fe1\u53f7 \u2265 +1.5\u03c3<\/td>\u6b63\u5411\u914d\u7f6e<\/td>\u8f83\u9ad8\u914d\u7f6e\u6c34\u5e73<\/td>\u4e2d\u7b49\u98ce\u9669<\/td><\/tr>
                  \u4fe1\u53f7 \u2264 -1.5\u03c3<\/td>\u53cd\u5411\u914d\u7f6e<\/td>\u7fa4\u4f53\u89c4\u907f\u914d\u7f6e<\/td>\u9ad8\u98ce\u9669<\/td><\/tr>
                  |\u4fe1\u53f7| < 1\u03c3<\/td>\u4e2d\u6027<\/td>\u7b49\u5f85\u4fe1\u53f7\u4e00\u81f4<\/td>\u4f4e\u98ce\u9669<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
                  \u5b9e\u65f6\u4fe1\u53f7\u751f\u6210\u4ee3\u7801<\/strong><\/h3>\n\n\n\n
                  # ------------------------------------------------<\/code># \u5b9e\u65f6\u4fe1\u53f7\u751f\u6210\uff1a\u57fa\u4e8e HBI\u3001CSAD\u3001\u7f51\u7edc\u5bc6\u5ea6\u6784\u5efa\u7efc\u5408\u914d\u7f6e\u4fe1\u53f7<\/code># ------------------------------------------------<\/code><\/code># \u521b\u5efa\u4fe1\u53f7DataFrame\uff0c\u4f7f\u7528\u5171\u540c\u7684\u65f6\u95f4\u7d22\u5f15<\/code>signals = pd.DataFrame(index=common_index)<\/code><\/code># ---------- 1. \u5355\u56e0\u5b50\u4fe1\u53f7 ----------<\/code># \u4fe1\u53f71\uff1aHBI\u504f\u79bb\uff08\u7f8a\u7fa4\u884c\u4e3a\u8fc7\u70ed\uff09<\/code>#        \u5f5360\u65e5HBI > 2.0\u65f6\uff0c\u8868\u660e\u4e2a\u80a1\u4e0e\u5e02\u573a\u80cc\u79bb\u4e25\u91cd\uff0c\u53ef\u80fd\u5b58\u5728\u8fc7\u5ea6\u7f8a\u7fa4<\/code>HBI_THRESHOLD = 2.0<\/code>signals[\"hbi_\u504f\u79bb\"] = (hbi_dict[\"HBI_60d\"] > HBI_THRESHOLD).astype(int)<\/code><\/code># \u4fe1\u53f72\uff1aCSAD\u96c6\u4e2d\uff08\u5e02\u573a\u540c\u6da8\u540c\u8dcc\uff09<\/code>#        \u5f53\u77ed\u671fCSAD\u4f4e\u4e8e5%\u5206\u4f4d\u6570\u65f6\uff0c\u8868\u660e\u4e2a\u80a1\u9ad8\u5ea6\u8d8b\u540c\uff0c\u5e02\u573a\u96c6\u4e2d\u5ea6\u9ad8<\/code>CSAD_QUANTILE = 0.05<\/code>signals[\"csad_\u96c6\u4e2d\"] = (csad_short < csad_short.quantile(CSAD_QUANTILE)).astype(int)<\/code><\/code># \u4fe1\u53f73\uff1a\u7f51\u7edc\u5bc6\u5ea6\u4f4e\uff08\u4e2a\u80a1\u95f4\u8054\u52a8\u5f31\uff09<\/code>#        \u5f53\u7f51\u7edc\u5bc6\u5ea6 < 0.5\u65f6\uff0c\u8868\u660e\u80a1\u7968\u95f4\u76f8\u5173\u6027\u8f83\u5f31\uff0c\u5e02\u573a\u7ed3\u6784\u677e\u6563<\/code>DENSITY_THRESHOLD = 0.5<\/code>signals[\"\u5bc6\u5ea6_\u4f4e\"] = (density_series < DENSITY_THRESHOLD).astype(int)<\/code><\/code># ---------- 2. \u7efc\u5408\u4fe1\u53f7 ----------<\/code># \u5bf9\u4e09\u4e2a\u5b50\u4fe1\u53f7\u6c42\u548c\uff0c\u5f97\u5230\u7efc\u5408\u5f97\u5206\uff08\u8303\u56f4\uff1a0 ~ 3\uff09<\/code>signals[\"\u7efc\u5408\"] = signals.sum(axis=1)<\/code><\/code># ---------- 3. \u6700\u7ec8\u914d\u7f6e\u4fe1\u53f7 ----------<\/code># \u7efc\u5408\u5f97\u5206 >= 2  \u2192 \u505a\u591a\uff08\u914d\u7f6e = 1\uff09<\/code># \u7efc\u5408\u5f97\u5206 <= 1  \u2192 \u505a\u7a7a\uff08\u914d\u7f6e = -1\uff09<\/code># \u7efc\u5408\u5f97\u5206 = 0   \u2192 \u7a7a\u4ed3\uff08\u914d\u7f6e = 0\uff09<\/code>signals[\"\u914d\u7f6e\"] = np.where(<\/code>    signals[\"\u7efc\u5408\"] >= 2,       # \u6761\u4ef6\uff1a\u81f3\u5c112\u4e2a\u4fe1\u53f7\u89e6\u53d1<\/code>    1,                          # \u505a\u591a<\/code>    np.where(<\/code>        signals[\"\u7efc\u5408\"] <= 1,   # \u6761\u4ef6\uff1a\u81f3\u591a1\u4e2a\u4fe1\u53f7\u89e6\u53d1<\/code>        -1,                     # \u505a\u7a7a<\/code>        0                       # \u7a7a\u4ed3\uff08\u7efc\u5408\u5f97\u5206\u4e3a0\u65f6\uff09<\/code>    )<\/code>)<\/code><\/code># ---------- \u53ef\u9009\uff1a\u67e5\u770b\u4fe1\u53f7\u5206\u5e03 ----------<\/code>print(\"\u4fe1\u53f7\u89e6\u53d1\u7edf\u8ba1\uff1a\")<\/code>print(signals[[\"hbi_\u504f\u79bb\", \"csad_\u96c6\u4e2d\", \"\u5bc6\u5ea6_\u4f4e\", \"\u7efc\u5408\", \"\u914d\u7f6e\"]].describe())<\/code><\/code><\/pre>\n\n\n\n
                  \"\"<\/figure>\n\n\n\n
                  \u751f\u4ea7\u7ea7\u4fe1\u53f7\u6821\u51c6\u77e9\u9635 – \u4f18\u5316\u53c2\u6570\u914d\u7f6e<\/p>\n\n\n\n
                  \u4e09\u3001\u5b9e\u6218\u6027\u80fd\u8868\u73b0\u5206\u6790\uff082018-2026\uff09<\/strong><\/h2>\n\n\n\n
                  \"\"<\/figure>\n\n\n\n
                  \u6027\u80fd\u6307\u6807<\/th>\u7b56\u7565\u8868\u73b0<\/th>\u82f1\u4f1f\u8fbe\u4e70\u5165\u6301\u6709<\/th>\u8d85\u989d\u6536\u76ca<\/th>\u6539\u5584\u5e45\u5ea6<\/th><\/tr>
                  \u5e74\u5316\u590d\u5408\u589e\u957f\u7387<\/td>47.2%<\/td>38.9%<\/td>+8.3%<\/td>+21.3%<\/td><\/tr>
                  \u6700\u5927\u56de\u64a4<\/td>-22%<\/td>-41%<\/td>+19%<\/td>+46.3%<\/td><\/tr>
                  \u590f\u666e\u6bd4\u7387<\/td>1.42<\/td>1.18<\/td>+0.24<\/td>+20.3%<\/td><\/tr>
                  \u80dc\u7387<\/td>67%<\/td>–<\/td>–<\/td>–<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
                  \u56db\u3001\u5b8c\u6574\u5468\u671f\u4ea4\u6613\u6846\u67b6\u4e0e\u5b9e\u6218\u6848\u4f8b<\/strong><\/h2>\n\n\n\n

                  \u5b8c\u6574\u5e02\u573a\u5468\u671f\uff086-12\u4e2a\u6708\uff09<\/strong><\/h3>\n\n\n\n
                    \n
                  1. CSAD\u89e6\u5e95 + \u5bc6\u5ea6\u4e0a\u5347<\/strong>\u00a0\u2192 \u964d\u4f4e\u6838\u5fc3\u98ce\u9669\u655e\u53e3;<\/li>\n\n\n\n
                  2. HBI\u6fc0\u589e\u504f\u79bb<\/strong>\u00a0\u2192 \u6218\u672f\u6027\u6b63\u5411\u914d\u7f6e;<\/li>\n\n\n\n
                  3. Q05\u56de\u5f52\u7cfb\u6570\u6076\u5316<\/strong>\u00a0\u2192 \u52a8\u6001\u5c3e\u90e8\u5bf9\u51b2;<\/li>\n\n\n\n
                  4. \u7f51\u7edc\u5206\u6563 + CSAD\u6fc0\u589e<\/strong>\u00a0\u2192 \u5168\u9762\u91cd\u65b0\u914d\u7f6e\u3002<\/li>\n<\/ol>\n\n\n\n
                    \ud83c\udfaf 2022\u5e74\u79d1\u6280\u80a1\u5468\u671f\u5b9e\u6218\u6848\u4f8b\uff1a<\/strong><\/p>\n\n\n\n
                    1\u6708<\/strong>\uff1aCSAD\u7b2c4\u767e\u5206\u4f4d + \u5bc6\u5ea668% \u2192 \u964d\u4f4e\u79d1\u6280\u80a1\u98ce\u9669\u655e\u53e3\uff1b
                    3\u6708<\/strong>\uff1a\u82f1\u4f1f\u8fbeHBI 3.2 \u2192 \u6218\u672f\u6027\u8d85\u914d\uff1b
                    6\u6708<\/strong>\uff1aQ05\u56de\u5f52\u7cfb\u6570-2.8 + \u5bc6\u5ea682% \u2192 \u9000\u51fa\u6240\u6709\u79d1\u6280\u80a1\u914d\u7f6e\uff1b
                    \u6700\u7ec8\u7ed3\u679c<\/strong>\uff1a\u6210\u529f\u907f\u514d40%\u56de\u64a4\u3002<\/p>\n\n\n\n
                    \u4e94\u3001\u65b9\u6cd5\u8bba\u5c40\u9650\u6027\u53ca\u7a33\u5065\u6027\u63aa\u65bd<\/strong><\/h2>\n\n\n\n

                    \u5173\u952e\u6df7\u6dc6\u56e0\u7d20\u8bc6\u522b<\/strong><\/h3>\n\n\n\n