{"id":2268,"date":"2025-05-12T07:20:00","date_gmt":"2025-05-11T23:20:00","guid":{"rendered":"https:\/\/blog.laoyulaoyu.top\/?p=2268"},"modified":"2025-05-03T15:50:27","modified_gmt":"2025-05-03T07:50:27","slug":"%e7%94%a8ai%e8%bd%bb%e6%9d%be%e8%af%86%e5%88%ab%e6%af%94%e7%89%b9%e5%b8%81%e5%a4%a7%e8%a1%8c%e6%83%85%ef%bc%81%e7%8b%ac%e5%ae%b6%e5%bc%80%e6%ba%9073%e5%87%86%e7%a1%ae%e7%8e%87%e7%9a%84%e9%a2%84","status":"publish","type":"post","link":"https:\/\/laoyulaoyu.com\/index.php\/2025\/05\/12\/%e7%94%a8ai%e8%bd%bb%e6%9d%be%e8%af%86%e5%88%ab%e6%af%94%e7%89%b9%e5%b8%81%e5%a4%a7%e8%a1%8c%e6%83%85%ef%bc%81%e7%8b%ac%e5%ae%b6%e5%bc%80%e6%ba%9073%e5%87%86%e7%a1%ae%e7%8e%87%e7%9a%84%e9%a2%84\/","title":{"rendered":"\u7528AI\u8f7b\u677e\u8bc6\u522b\u6bd4\u7279\u5e01\u5927\u884c\u60c5\uff01\u72ec\u5bb6\u5f00\u6e9073%\u51c6\u786e\u7387\u7684\u9884\u6d4b\u6a21\u578b"},"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
\u5199\u5728\u524d\u9762\u7684\u8bdd\uff1a<\/strong>\u6211\u662f\u8001\u4f59\u635e\u9c7c\uff0c\u4eca\u5929\u624b\u628a\u624b\u6559\u4f60\u7528AI\u9884\u6d4b\u6bd4\u7279\u5e01\u3001\u4ee5\u592a\u574a\u7b49\u8d8b\u52bf\uff01\u5728\u7b14\u8bb0\u672c\u7535\u8111\u4e0a\u5c31\u80fd\u8bad\u7ec3\u4e00\u4e2aLSTM\u6a21\u578b\uff0c\u63d0\u524d\u5224\u65ad\u4ef7\u683c\u60ef\u6027\u65b9\u5411\uff0c\u5b9e\u6d4b\u51c6\u786e\u738774%\u3002<\/mark>\u54ea\u6015\u4f60\u662f\u5c0f\u767d\uff0c\u770b\u5b8c\u4e5f\u80fd\u5b66\u4f1a\u8fd9\u62db\u8fa8\u725b\u718a\u3001\u6293\u8d8b\u52bf\u7684AI\u91cf\u5316\u795e\u5668\uff01<\/pre>\n<\/blockquote>\n\n\n\n
\u4e00\u3001\u81ea\u76f8\u5173\u6027\uff1a\u6bd4\u7279\u5e01\u7684\u201c\u60ef\u6027\u5bc6\u7801\u201d<\/strong><\/h2>\n\n\n\n
1. \u4ec0\u4e48\u662f\u81ea\u76f8\u5173\u6027\uff1f<\/strong><\/p>\n\n\n\n
\u7b80\u5355\u8bf4\uff0c\u5c31\u662f\u201c\u6628\u5929\u7684\u6da8\u8dcc\u5982\u4f55\u5f71\u54cd\u4eca\u5929\u201d\u3002\u6bd4\u5982\uff1a<\/p>\n\n\n\n
    \n
  • \u81ea\u76f8\u5173\u7cfb\u6570 >0<\/strong>\uff1a\u6628\u5929\u6da8\uff0c\u4eca\u5929\u5927\u6982\u7387\u7ee7\u7eed\u6da8\uff08\u725b\u5e02\u7279\u5f81\uff09<\/li>\n\n\n\n
  • \u81ea\u76f8\u5173\u7cfb\u6570 \u22480<\/strong>\uff1a\u4ef7\u683c\u968f\u673a\u6e38\u8d70\uff08\u718a\u5e02\u9707\u8361\uff09<\/li>\n\n\n\n
  • \u81ea\u76f8\u5173\u7cfb\u6570 <0<\/strong>\uff1a\u6628\u5929\u6da8\uff0c\u4eca\u5929\u53ef\u80fd\u8dcc\uff08\u5e84\u5bb6\u6d17\u76d8\uff09<\/li>\n<\/ul>\n\n\n\n
    \"\"<\/figure>\n\n\n\n
    2. \u4e3a\u4ec0\u4e48\u8981\u9884\u6d4b\u5b83\uff1f<\/strong><\/p>\n\n\n\n
    \u666e\u901a\u6563\u6237\u770bK\u7ebf\uff0c\u9ad8\u624b\u770b\u5e95\u5c42\u89c4\u5f8b\uff01\u901a\u8fc7\u9884\u6d4b\u81ea\u76f8\u5173\u6027\u53d8\u5316\uff0c\u4f60\u80fd\uff1a
    \u2705 \u8bc6\u522b\u8d8b\u52bf\u542f\u52a8\uff08\u5982\u7cfb\u6570\u4ece0.3\u98d9\u5347\u52300.8\uff09
    \u2705 \u907f\u5f00\u9707\u8361\u9677\u9631\uff08\u7cfb\u6570\u6301\u7eed\u4f4e\u4e8e0.2\u65f6\u8eba\u5e73\uff09
    \u2705 \u9884\u5224\u5e84\u5bb6\u52a8\u4f5c\uff08\u7cfb\u6570\u7a81\u7136\u8f6c\u8d1f\u53ef\u80fd\u662f\u7838\u76d8\u4fe1\u53f7\uff09<\/p>\n\n\n\n
    \u4e8c\u3001\u4ee3\u7801\u5b9e\u6218\uff1a\u4ece\u6570\u636e\u5230\u9884\u6d4b\u5168\u6d41\u7a0b<\/strong><\/h2>\n\n\n\n
    \"\"<\/figure>\n\n\n\n
    \u258d\u6b65\u9aa41\uff1a\u83b7\u53d6\u6bd4\u7279\u5e01\u6570\u636e<\/strong><\/p>\n\n\n\n
    \u7528yfinance<\/code>\u5e93\u76f4\u63a5\u8c03\u53d6\u6700\u65b0\u6570\u636e\uff0c\u8fde\u4e0aVPN\u7684\u529f\u592b\u90fd\u7701\u4e86\uff1a<\/p>\n\n\n\n
    # 1. \u6570\u636e\u7ba1\u9053 -----------------------------------------------
    print(\"\ud83d\ude80 \u6b63\u5728\u83b7\u53d6\u6bd4\u7279\u5e01\u6570\u636e...\")
    data = yf.download('BTC-USD', 
                      start='2020-01-01',
                      end=pd.Timestamp.today().strftime('%Y-%m-%d'),
                      progress=False)['Close']<\/pre>\n\n\n\n
    \u26a0\ufe0f \u5751\u70b9\u63d0\u9192<\/strong>\uff1a<\/p>\n\n\n\n
      \n
    • \u5982\u679c\u6570\u636e\u5305\u542bNaN\uff0c\u7528btc_data = btc_data.dropna()<\/code>\u6e05\u7406<\/li>\n\n\n\n
    • \u56fd\u5185\u7528\u6237\u53ef\u80fd\u9700\u5728\u4ee3\u7801\u524d\u52a0import pandas as pd<\/code><\/li>\n<\/ul>\n\n\n\n
      \u258d\u6b65\u9aa42\uff1a\u8ba1\u7b97\u6eda\u52a8\u81ea\u76f8\u5173<\/strong><\/p>\n\n\n\n
      \u7528Pandas\u7684.rolling()<\/code>\u51fd\u6570\uff0c\u4e00\u952e\u751f\u621030\u5929\u60ef\u6027\u6307\u6807\uff1a<\/p>\n\n\n\n
      print(\"\ud83e\uddee \u8ba1\u7b97\u6eda\u52a8\u81ea\u76f8\u5173...\")
      rolling_autocorr = data.rolling(window=30).apply(
          lambda x: x.autocorr(lag=1), 
          raw=False
      ).dropna().values
      print(f\"\u2705 \u6570\u636e\u5c31\u7eea\uff01\u6709\u6548\u6837\u672c\u6570\uff1a{len(rolling_autocorr)}\") <\/pre>\n\n\n\n
      \ud83d\udd0d \u539f\u7406\u56fe\u89e3<\/strong>\uff1a<\/p>\n\n\n\n
      \u65f6\u95f4\u7a97\u53e3<\/th>\u8ba1\u7b97\u65b9\u5f0f<\/th><\/tr><\/thead>
      \u7b2c1-30\u5929<\/td>\u8ba1\u7b97\u7b2c1\u5929\u4e0e\u7b2c2\u5929\u7684\u76f8\u5173\u6027<\/td><\/tr>
      \u7b2c2-31\u5929<\/td>\u8ba1\u7b97\u7b2c2\u5929\u4e0e\u7b2c3\u5929\u7684\u76f8\u5173\u6027<\/td><\/tr>
      …<\/td>…<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
      \u258d\u6b65\u9aa43\uff1a\u6570\u636e\u683c\u5f0f\u5316\u2014\u2014LSTM\u7684\u201c\u4e13\u5c5e\u5957\u9910\u201d<\/strong><\/p>\n\n\n\n
      LSTM\u6a21\u578b\u53ea\u5403\u201c\u65f6\u95f4\u5e8f\u5217\u5957\u9910\u201d\uff0c\u9700\u628a\u6570\u636e\u5207\u621090\u5929\u4e00\u6bb5\u7684\u5386\u53f2\u7247\u6bb5\uff1a<\/p>\n\n\n\n
      def create_dataset(data, n_steps=90):
          \"\"\"\u5c06\u65f6\u95f4\u5e8f\u5217\u8f6c\u6362\u4e3a\u76d1\u7763\u5b66\u4e60\u683c\u5f0f\"\"\"
          X, y = [], []
          for i in range(n_steps, len(data)):
              X.append(data[i-n_steps:i])
              y.append(data[i])
          return np.array(X), np.array(y)

      N_STEPS = 90
      X, y = create_dataset(rolling_autocorr, N_STEPS)
      X = X.reshape((X.shape[0], N_STEPS, 1))

      # \u6570\u636e\u5206\u5272
      split = int(0.8 * len(X))
      X_train, X_test = X[:split], X[split:]
      y_train, y_test = y[:split].flatten(), y[split:].flatten()<\/pre>\n\n\n\n
      \ud83d\udea8 \u65b0\u624b\u5fc5\u770b<\/strong>\uff1a<\/p>\n\n\n\n
        \n
      • \u5982\u679c\u62a5\u9519ValueError<\/code>\uff0c\u68c0\u67e5\u6570\u636e\u662f\u5426\u5305\u542b\u975e\u6570\u503c\uff08\u5982\u6587\u672c\uff09<\/li>\n\n\n\n
      • \u6570\u636e\u6807\u51c6\u5316\u53ef\u63d0\u5347\u8bad\u7ec3\u901f\u5ea6\uff08\u4ee3\u7801\u89c1\u6587\u672b\u6e90\u7801\uff09<\/li>\n<\/ul>\n\n\n\n
        \u4e09\u3001\u6a21\u578b\u8bad\u7ec3\uff1a\u4e09\u4e2a\u6280\u5de7\u8ba9AI\u66f4\u806a\u660e<\/strong><\/h2>\n\n\n\n
        \u6280\u5de71\uff1aLSTM\u5c42\u2014\u2014\u8bb0\u5fc6\u5927\u5e08<\/strong><\/p>\n\n\n\n