\u4e0a\u4e00\u8bb2\u6211\u4eec\u642d\u5efa\u4e86Python\u91cf\u5316\u5f00\u53d1\u73af\u5883，\u5e76\u901a\u8fc7akshare\u5b8c\u6210\u4e86\u7b2c\u4e00\u6b21"\u4ece\u6570\u636e\u5230\u51b3\u7b56"\u7684\u591a\u56e0\u5b50\u9009\u80a1\u5b9e\u9a8c\u3002\u4f46\u8fd9\u91cc\u6709\u4e00\u4e2a\u9690\u60a3\u4f60\u53ef\u80fd\u5df2\u7ecf\u5bdf\u89c9\u4e86：\u6bcf\u6b21\u8fd0\u884c\u7b56\u7565\u90fd\u9700\u8981\u4e34\u65f6\u4ece\u7f51\u4e0a\u6293\u6570\u636e\u2014\u2014\u6162\u3001\u4e0d\u7a33\u5b9a\u3001\u4e0d\u53ef\u590d\u73b0\u3002<\/p>\n
\u771f\u6b63\u80fd\u6253\u4ed7\u7684\u91cf\u5316\u56e2\u961f，\u7edd\u4e0d\u4f1a\u6bcf\u8dd1\u4e00\u6b21\u56de\u6d4b\u5c31\u53bb\u7f51\u4e0a\u91cd\u65b0\u4e0b\u8f7d\u4e00\u6b21\u6570\u636e\u3002\u4ed6\u4eec\u628a\u6570\u636e\u91c7\u96c6\u3001\u6e05\u6d17\u3001\u5b58\u50a8\u505a\u6210\u4e00\u6761\u81ea\u52a8\u5316\u7ba1\u9053（Pipeline），\u9700\u8981\u65f6\u4ece\u672c\u5730\u6570\u636e\u5e93\u79d2\u7ea7\u8bfb\u53d6\u3002\u8fd9\u6761\u6570\u636e\u7ba1\u9053\u5c31\u662f\u91cf\u5316\u4ea4\u6613\u7684"\u5f39\u836f\u5e93"\u2014\u2014\u6ca1\u6709\u7a33\u5b9a\u7684\u6570\u636e\u4f9b\u7ed9，\u7b56\u7565\u7814\u7a76\u5c31\u662f\u7eb8\u4e0a\u8c08\u5175\u3002<\/p>\n
\u4eca\u5929，\u5c0f\u6674（\u60c5\u62a5\u5b98）\u6b63\u5f0f\u4e0a\u5c97\u3002\u4ed6\u5c06\u8d1f\u8d23\u628a\u5206\u6563\u5728\u5404\u4e2a\u6570\u636e\u6e90\u7684\u5e02\u573a\u4fe1\u606f\u2014\u2014\u884c\u60c5\u3001\u8d22\u52a1\u3001\u5b8f\u89c2\u3001\u7814\u62a5\u3001\u65b0\u95fb\u2014\u2014\u7edf\u4e00\u6536\u62e2\u3001\u6e05\u6d17\u3001\u6807\u51c6\u5316\u540e\u5b58\u5165\u672c\u5730\u6570\u636e\u5e93，\u5e76\u5728\u540e\u7eed\u8bfe\u7a0b\u4e2d\u9010\u6b65\u8fdb\u9636\u4e3a\u80fd\u9605\u8bfb\u7814\u62a5\u3001\u611f\u77e5\u8206\u60c5\u7684AI\u6295\u7814\u4e13\u5bb6\u3002<\/p>\n
\u5b66\u5b8c\u672c\u8282\u540e，\u4f60\u5c06\u80fd\u591f：<\/p>\n
\u5148\u770b\u4e00\u4e2a\u771f\u5b9e\u573a\u666f：\u4f60\u8981\u56de\u6d4b\u4e00\u4e2a\u57fa\u4e8e\u8d22\u52a1\u56e0\u5b50（PE\u3001ROE）\u548c\u6280\u672f\u6307\u6807（MACD\u3001RSI）\u7684\u7ec4\u5408\u7b56\u7565\u3002\u8fd9\u9700\u8981\u4e24\u7c7b\u6570\u636e：\u65e5\u7ebf\u884c\u60c5（\u7528\u6765\u7b97\u6280\u672f\u6307\u6807）\u548c\u5b63\u5ea6\u8d22\u52a1\u62a5\u544a（\u7528\u6765\u7b97\u4f30\u503c\u56e0\u5b50）\u3002<\/p>\n
\u5982\u679c\u6bcf\u6b21\u56de\u6d4b\u524d\u90fd\u4ece\u7f51\u4e0a\u4e0b\u8f7d\u2014\u2014akshare\u53d6\u884c\u60c5（\u7ea610\u79d2）\u3001\u53d6\u8d22\u52a1\u6570\u636e（\u7ea630\u79d2）\u3001\u53d6\u5b8f\u89c2\u6307\u6807（\u7ea615\u79d2）\u2014\u2014\u5149\u6570\u636e\u51c6\u5907\u5c31\u8981\u4e00\u5206\u949f\u3002\u4e00\u5929\u8fed\u4ee3\u6d4b\u8bd550\u4e2a\u53c2\u6570\u7ec4\u5408，\u7b49\u5f85\u65f6\u95f4\u5c31\u63a5\u8fd1\u4e00\u5c0f\u65f6\u4e86\u3002\u66f4\u81f4\u547d\u7684\u662f，\u7f51\u7edc\u6ce2\u52a8\u53ef\u80fd\u5bfc\u81f4\u67d0\u6b21\u4e0b\u8f7d\u5931\u8d25，\u4f60\u7684\u6279\u91cf\u56de\u6d4b\u811a\u672c\u5c31\u4f1a\u5361\u5728\u534a\u8def，\u4e00\u884c\u4e00\u884c\u6392\u67e5\u9519\u8bef\u3002<\/p>\n
\u5c0f\u6674\u7684\u6570\u636e\u7ba1\u9053\u8981\u89e3\u51b3\u7684\u5c31\u662f\u8fd9\u4e2a\u95ee\u9898\u3002\u5b83\u7684\u8bbe\u8ba1\u67b6\u6784\u5982\u4e0b：<\/p>\n
\u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510 \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510 \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510
\n\u2502 \u6570\u636e\u6e90A \u2502 \u2502 \u6570\u636e\u6e90B \u2502 \u2502 \u6570\u636e\u6e90C \u2502
\n\u2502 akshare \u2502 \u2502 tushare \u2502 \u2502 QMT\u884c\u60c5 \u2502
\n\u2514\u2500\u2500\u2500\u2500\u252c\u2500\u2500\u2500\u2500\u2500\u2518 \u2514\u2500\u2500\u2500\u2500\u252c\u2500\u2500\u2500\u2500\u2500\u2518 \u2514\u2500\u2500\u2500\u2500\u252c\u2500\u2500\u2500\u2500\u2500\u2518
\n \u2502 \u2502 \u2502
\n \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u253c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518
\n \u25bc
\n \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510
\n \u2502 \u5c0f\u6674\u91c7\u96c6\u5668 \u2502 \u2190 \u5b9a\u65f6\/\u624b\u52a8\u89e6\u53d1
\n \u2502 \u591a\u6e90\u5e76\u884c\u62c9\u53d6 \u2502
\n \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u252c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518
\n \u25bc
\n \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510
\n \u2502 \u6e05\u6d17\u4e0e\u6807\u51c6\u5316 \u2502 \u2190 \u590d\u6743\u5904\u7406\u3001\u505c\u724c\u6807\u8bb0\u3001\u53bb\u91cd
\n \u2502 \u7f3a\u5931\u503c\u586b\u8865 \u2502
\n \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u252c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518
\n \u25bc
\n \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510
\n \u2502 \u672c\u5730\u6570\u636e\u5e93 \u2502 \u2190 HDF5 \/ CSV \/ SQLite
\n \u2502 (\u7ed3\u6784\u5316\u5b58\u50a8) \u2502
\n \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u252c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518
\n \u25bc
\n \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510
\n \u2502 \u5c0f\u96c5\u7b56\u7565\u56de\u6d4b \u2502 \u2190 \u79d2\u7ea7\u8bfb\u53d6，\u65e0\u9700\u8054\u7f51
\n \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518<\/p>\n
\u6838\u5fc3\u8bbe\u8ba1\u539f\u5219\u6709\u4e09\u6761：<\/p>\n
\u4e0b\u9762\u662f\u5c0f\u6674\u6570\u636e\u7ba1\u9053\u7684\u6838\u5fc3\u91c7\u96c6\u5668\u6846\u67b6：<\/p>\n
"""
\n\u5c0f\u6674\u6570\u636e\u7ba1\u9053 – \u6838\u5fc3\u91c7\u96c6\u6846\u67b6
\n\u804c\u8d23：\u7edf\u4e00\u7ba1\u7406\u6570\u636e\u91c7\u96c6\u3001\u589e\u91cf\u66f4\u65b0\u3001\u672c\u5730\u5b58\u50a8
\n"""
\nimport akshare as ak
\nimport pandas as pd
\nimport os
\nfrom datetime import datetime, timedelta<\/p>\n
class XiaoQingDataPipeline:
\n """\u5c0f\u6674\u60c5\u62a5\u5b98\u7684\u6570\u636e\u7ba1\u9053"""<\/p>\n
def __init__(self, data_dir=".\/local_data"):
\n """\u521d\u59cb\u5316：\u8bbe\u7f6e\u672c\u5730\u5b58\u50a8\u76ee\u5f55，\u521b\u5efa\u5b50\u6587\u4ef6\u5939"""
\n self.data_dir = data_dir
\n self.daily_dir = os.path.join(data_dir, "daily") # \u65e5\u7ebf\u884c\u60c5
\n self.finance_dir = os.path.join(data_dir, "finance") # \u8d22\u52a1\u6570\u636e
\n self.macro_dir = os.path.join(data_dir, "macro") # \u5b8f\u89c2\u6570\u636e
\n for d in [self.daily_dir, self.finance_dir, self.macro_dir]:
\n os.makedirs(d, exist_ok=True)<\/p>\n
# — \u6a21\u5757\u4e00：\u65e5\u7ebf\u884c\u60c5\u91c7\u96c6 —
\n def fetch_daily_kline(self, symbol, start_date="20200101",
\n end_date=None, force_refresh=False):
\n """
\n \u91c7\u96c6\u5355\u53ea\u80a1\u7968\u7684\u65e5K\u7ebf\u6570\u636e，\u652f\u6301\u589e\u91cf\u66f4\u65b0<\/p>\n
\u53c2\u6570：
\n symbol: \u80a1\u7968\u4ee3\u7801，\u5982 '600519'
\n start_date: \u8d77\u59cb\u65e5\u671f
\n end_date: \u7ed3\u675f\u65e5\u671f，\u9ed8\u8ba4\u4e3a\u4eca\u5929
\n force_refresh: \u662f\u5426\u5f3a\u5236\u5168\u91cf\u66f4\u65b0
\n """
\n if end_date is None:
\n end_date = datetime.now().strftime("%Y%m%d")<\/p>\n
file_path = os.path.join(self.daily_dir, f"{symbol}.csv")<\/p>\n
# \u589e\u91cf\u66f4\u65b0\u903b\u8f91：\u5982\u679c\u672c\u5730\u5df2\u6709\u6570\u636e，\u53ea\u62c9\u53d6\u65b0\u589e\u90e8\u5206
\n if not force_refresh and os.path.exists(file_path):
\n existing = pd.read_csv(file_path, index_col=0, parse_dates=True)
\n last_date = existing.index.max().strftime("%Y%m%d")
\n if last_date >= end_date:
\n print(f"[{symbol}] \u6570\u636e\u5df2\u662f\u6700\u65b0，\u8df3\u8fc7")
\n return existing
\n # \u4ece\u6700\u540e\u65e5\u671f+1\u5929\u5f00\u59cb\u62c9\u53d6
\n start_date = (pd.to_datetime(last_date) +
\n timedelta(days=1)).strftime("%Y%m%d")
\n print(f"[{symbol}] \u589e\u91cf\u66f4\u65b0：{start_date} \u2192 {end_date}")<\/p>\n
# \u4eceakshare\u62c9\u53d6\u6570\u636e
\n df = ak.stock_zh_a_hist(symbol=symbol, period="daily",
\n start_date=start_date, end_date=end_date,
\n adjust="qfq")
\n df['\u65e5\u671f'] = pd.to_datetime(df['\u65e5\u671f'])
\n df.set_index('\u65e5\u671f', inplace=True)<\/p>\n
# \u5982\u679c\u6709\u5386\u53f2\u6570\u636e，\u5408\u5e76
\n if os.path.exists(file_path) and not force_refresh:
\n existing = pd.read_csv(file_path, index_col=0, parse_dates=True)
\n df = pd.concat([existing, df])
\n df = df[~df.index.duplicated(keep='last')] # \u53bb\u91cd，\u4fdd\u7559\u6700\u65b0<\/p>\n
# \u5b58\u5165\u672c\u5730CSV
\n df.to_csv(file_path)
\n print(f"[{symbol}] \u4fdd\u5b58\u5b8c\u6210，\u5171 {len(df)} \u6761\u8bb0\u5f55")
\n return df<\/p>\n
# — \u6a21\u5757\u4e8c：\u5b8f\u89c2\u6570\u636e\u91c7\u96c6 —
\n def fetch_macro_data(self):
\n """\u91c7\u96c6\u5173\u952e\u5b8f\u89c2\u7ecf\u6d4e\u6307\u6807"""
\n indicators = {}<\/p>\n
# \u83b7\u53d6\u8d27\u5e01\u4f9b\u5e94\u91cf\u6570\u636e（M1\/M2\u662f\u8861\u91cf\u5e02\u573a\u6d41\u52a8\u6027\u7684\u5173\u952e\u6307\u6807）
\n try:
\n money_supply = ak.macro_china_money_supply()
\n indicators['money_supply'] = money_supply
\n print("\u5b8f\u89c2\u6570\u636e：\u8d27\u5e01\u4f9b\u5e94\u91cf\u91c7\u96c6\u5b8c\u6210")
\n except Exception as e:
\n print(f"\u5b8f\u89c2\u6570\u636e\u91c7\u96c6\u5931\u8d25（\u8d27\u5e01\u4f9b\u5e94\u91cf）：{e}")<\/p>\n
# \u83b7\u53d6CPI\u6570\u636e（\u901a\u8d27\u81a8\u80c0\u7387\u6307\u6807）
\n try:
\n cpi = ak.macro_china_cpi_monthly()
\n indicators['cpi'] = cpi
\n print("\u5b8f\u89c2\u6570\u636e：CPI\u91c7\u96c6\u5b8c\u6210")
\n except Exception as e:
\n print(f"\u5b8f\u89c2\u6570\u636e\u91c7\u96c6\u5931\u8d25（CPI）：{e}")<\/p>\n
# \u4fdd\u5b58\u5230\u672c\u5730
\n for name, data in indicators.items():
\n file_path = os.path.join(self.macro_dir, f"{name}.csv")
\n data.to_csv(file_path, index=False)<\/p>\n
return indicators<\/p>\n
# — \u6a21\u5757\u4e09：\u8d22\u52a1\u6570\u636e\u91c7\u96c6 —
\n def fetch_financial_data(self, symbol, indicator="\u6309\u62a5\u544a\u671f"):
\n """
\n \u91c7\u96c6\u4e2a\u80a1\u8d22\u52a1\u6307\u6807（\u4ee5\u540c\u82b1\u987a\u63a5\u53e3\u4e3a\u4f8b）<\/p>\n
\u8fd4\u56de\u503c\u5305\u542b：\u6bcf\u80a1\u6536\u76ca\u3001\u6bcf\u80a1\u51c0\u8d44\u4ea7\u3001\u51c0\u8d44\u4ea7\u6536\u76ca\u7387\u3001\u6bdb\u5229\u7387\u7b49
\n """
\n try:
\n df = ak.stock_financial_abstract_ths(symbol=symbol,
\n indicator=indicator)
\n file_path = os.path.join(self.finance_dir, f"{symbol}_finance.csv")
\n df.to_csv(file_path, index=False)
\n print(f"[{symbol}] \u8d22\u52a1\u6570\u636e\u91c7\u96c6\u5b8c\u6210，\u5171 {len(df)} \u6761")
\n return df
\n except Exception as e:
\n print(f"[{symbol}] \u8d22\u52a1\u6570\u636e\u91c7\u96c6\u5931\u8d25：{e}")
\n return None<\/p>\n
# === \u4f7f\u7528\u793a\u4f8b：\u8ba9\u5c0f\u6674\u8dd1\u4e00\u6b21\u5168\u91cf\u91c7\u96c6 ===
\nxiaoQing = XiaoQingDataPipeline(data_dir=".\/local_data")<\/p>\n
# \u91c7\u96c6\u80a1\u7968\u6c60\u7684\u65e5\u7ebf\u6570\u636e
\nstock_pool = ['600519', '000001', '000858', '600036']
\nfor code in stock_pool:
\n xiaoQing.fetch_daily_kline(code, start_date="20240101")<\/p>\n
\u4ece\u5916\u90e8\u6570\u636e\u6e90\u83b7\u53d6\u7684\u539f\u59cb\u6570\u636e\u51e0\u4e4e100%\u5b58\u5728\u8d28\u91cf\u95ee\u9898\u3002\u4ee5\u4e0b\u662f\u91cf\u5316\u6570\u636e\u4e2d\u6700\u5e38\u89c1\u7684\u56db\u7c7b\u95ee\u9898\u53ca\u5176\u5904\u7406\u65b9\u6848：<\/p>\n
| \u590d\u6743\u7f3a\u5931<\/td>\n | \u5386\u53f2\u4ef7\u683c\u5728\u9664\u6743\u9664\u606f\u65e5\u51fa\u73b0"\u65ad\u5d16\u5f0f"\u8df3\u7a7a<\/td>\n | \u7edf\u4e00\u4f7f\u7528\u524d\u590d\u6743（qfq），\u907f\u514d\u542b\u6743\u4ef7\u683c\u5f71\u54cd\u56de\u6d4b<\/td>\n<\/tr>\n | ||||||||||||
| \u505c\u724c\u7a7a\u503c<\/td>\n | \u4ea4\u6613\u65e5\u65e0\u6210\u4ea4，OHLC\u4e3aNaN\u6216\u91cd\u590d\u524d\u4e00\u65e5<\/td>\n | \u6807\u8bb0\u505c\u724c\u671f，\u56de\u6d4b\u4e2d\u8df3\u8fc7\u6216\u586b\u5145\u524d\u6536\u76d8\u4ef7<\/td>\n<\/tr>\n | ||||||||||||
| \u8d22\u52a1\u6570\u636e\u6ede\u540e<\/td>\n | 12\u670831\u65e5\u7684\u5e74\u62a5，\u5b9e\u9645\u6b21\u5e744\u6708\u624d\u516c\u5e03<\/td>\n | \u4f7f\u7528\u65f6\u70b9\u5339\u914d：4\u6708\u4e4b\u524d\u7528\u4e09\u5b63\u5ea6\u62a5，4\u6708\u4e4b\u540e\u53ef\u7528\u5e74\u62a5<\/td>\n<\/tr>\n | ||||||||||||
| \u5f02\u5e38\u6781\u503c<\/td>\n | \u9519\u8bef\u5f55\u5165\u5bfc\u81f4PE=99999\u6216ROE=500%<\/td>\n | \u8bbe\u5b9a\u5408\u7406\u9608\u503c，\u8d85\u8fc7\u9608\u503c\u7684\u6807\u8bb0\u4e3a\u7f3a\u5931\u5e76\u53e6\u884c\u5904\u7406<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n \u5176\u4e2d\u8d22\u52a1\u6570\u636e\u7684"\u65f6\u70b9\u5339\u914d"\u662f\u6700\u5bb9\u6613\u88ab\u5ffd\u89c6\u4f46\u5f71\u54cd\u6700\u5927\u7684\u95ee\u9898\u3002\u4f8b\u5982，2024\u5e741\u6708\u505a\u56de\u6d4b，\u4f60"\u770b\u5230"\u7684\u662f2023\u5e74\u5e74\u62a5\u7684\u6570\u636e（\u901a\u5e38\u57283-4\u6708\u624d\u516c\u5e03），\u8fd9\u5c5e\u4e8e"\u5077\u770b\u7b54\u6848"\u2014\u2014\u5b9e\u76d8\u4e2d1\u6708\u4efd\u6839\u672c\u4e0d\u77e5\u9053\u8fd9\u4e2a\u6570\u636e\u3002<\/p>\n \u4ee3\u7801\u6f14\u793a<\/h4>\n\u4e0b\u9762\u5b9e\u73b0\u4e00\u4e2a\u6570\u636e\u6e05\u6d17\u5668，\u81ea\u52a8\u5904\u7406\u4e0a\u8ff0\u56db\u7c7b\u95ee\u9898：<\/p>\n class DataCleaner: @staticmethod \u53c2\u6570： # 1. \u6807\u8bb0\u505c\u724c\u65e5：\u6210\u4ea4\u91cf=0 \u6216 \u6700\u9ad8\u4ef7=\u6700\u4f4e\u4ef7（\u4e00\u5b57\u505c\u724c） # 2. \u524d\u5411\u586b\u5145\u505c\u724c\u65e5\u7684\u6536\u76d8\u4ef7（\u7528\u4e8e\u8ba1\u7b97\u6536\u76ca\u7387\u65f6\u4fdd\u6301\u8fde\u7eed\u6027） # 3. \u68c0\u6d4b\u5f02\u5e38\u4ef7\u683c\u53d8\u52a8：\u5355\u65e5\u6da8\u8dcc\u5e45\u8d85\u8fc7\u6da8\u8dcc\u505c\u9650\u5236 if cleaned['is_abnormal'].sum() > 0: return cleaned<\/p>\n @staticmethod \u53c2\u6570： @staticmethod A\u80a1\u5e74\u62a5\u622a\u6b62\u65e5\u4e3a4\u670830\u65e5，\u4e2d\u62a5\u4e3a8\u670831\u65e5，\u5b63\u62a5\u4e3a\u5f53\u5b63\u672b\u6b21\u6708 def get_available_date(report_date): df['available_date'] = df[report_col].apply(get_available_date) \u5173\u952e\u8981\u70b9<\/h4>\n\n 4.3 \u50ac\u5316\u5242\u8bc6\u522b\u2014\u2014\u7528\u5927\u6a21\u578b\u8bfb\u61c2\u5e02\u573a\u4e8b\u4ef6<\/h3>\n |
\u9664\u4e86\u7ed3\u6784\u5316\u6570\u636e（\u4ef7\u683c\u3001\u8d22\u52a1），\u5e02\u573a\u8fd8\u5b58\u5728\u5927\u91cf\u975e\u7ed3\u6784\u5316\u4fe1\u606f\u2014\u2014\u65b0\u95fb\u516c\u544a\u3001\u653f\u7b56\u6587\u4ef6\u3001\u884c\u4e1a\u7814\u62a5\u3001\u793e\u4ea4\u5a92\u4f53\u8ba8\u8bba\u3002\u8fd9\u4e9b\u4fe1\u606f\u4e2d\u8574\u85cf\u7740\u53ef\u80fd\u7684"\u50ac\u5316\u5242\u4e8b\u4ef6"：<\/p>\n
\u4f20\u7edf\u65b9\u6cd5\u7528\u5173\u952e\u8bcd\u5339\u914d\u6765\u8bc6\u522b\u4e8b\u4ef6（\u5982"\u540c\u6bd4\u589e\u957f + \u51c0\u5229\u6da6 > 100%"），\u4f46\u4f1a\u6f0f\u6389\u5f88\u591a"\u4e0d\u50cf\u5173\u952e\u8bcd\u4f46\u786e\u5b9e\u662f\u4e8b\u4ef6"\u7684\u60c5\u51b5\u3002\u5927\u6a21\u578b\u5728\u8fd9\u4e00\u6b65\u5927\u6709\u7528\u6b66\u4e4b\u5730\u2014\u2014\u5b83\u80fd\u7406\u89e3\u53e5\u5b50\u7684\u8bed\u4e49，\u800c\u4e0d\u53ea\u662f\u626b\u63cf\u5173\u952e\u8bcd\u3002<\/p>\n
\u5c0f\u6674\u7684\u5de5\u4f5c\u6d41\u7a0b\u662f：<\/p>\n
\u4e0b\u9762\u662f\u4e00\u4e2a\u5229\u7528OpenAI API（\u6216\u5176\u4ed6\u517c\u5bb9\u63a5\u53e3）\u505a\u50ac\u5316\u5242\u4e8b\u4ef6\u8bc6\u522b\u7684\u65b9\u6cd5：<\/p>\n
"""
\n\u5c0f\u6674\u50ac\u5316\u5242\u4e8b\u4ef6\u8bc6\u522b\u6a21\u5757
\n\u2014\u2014 \u8ba9\u5927\u6a21\u578b\u5e2e\u4f60\u8bfb\u516c\u544a，\u63d0\u53d6\u7ed3\u6784\u5316\u4e8b\u4ef6\u4fe1\u606f
\n"""
\nimport json<\/p>\n
class CatalystDetector:
\n """\u50ac\u5316\u5242\u4e8b\u4ef6\u8bc6\u522b\u5668"""<\/p>\n
# \u5b9a\u4e49\u63d0\u793a\u8bcd\u6a21\u677f
\n DETECTION_PROMPT = """\u4f60\u662f\u4e00\u4e2aA\u80a1\u5e02\u573a\u4e8b\u4ef6\u5206\u6790\u4e13\u5bb6\u3002\u8bf7\u9605\u8bfb\u4ee5\u4e0b\u65b0\u95fb\/\u516c\u544a，\u5224\u65ad\u5b83\u662f\u5426\u662f\u4e00\u4e2a\u503c\u5f97\u91cf\u5316\u7b56\u7565\u5173\u6ce8\u7684"\u50ac\u5316\u5242\u4e8b\u4ef6"\u3002<\/p>\n
\u8f93\u51fa\u8981\u6c42：\u4e25\u683c\u8fd4\u56deJSON\u683c\u5f0f，\u5305\u542b\u4ee5\u4e0b\u5b57\u6bb5：
\n– is_catalyst: true\/false，\u662f\u5426\u6784\u6210\u50ac\u5316\u5242
\n– event_type: \u4e8b\u4ef6\u7c7b\u578b（\u8d85\u9884\u671f\u4e1a\u7ee9\/\u653f\u7b56\u5229\u597d\/\u91cd\u5927\u5408\u540c\/\u884c\u4e1a\u53d8\u9769\/\u98ce\u9669\u8b66\u793a\/\u5176\u4ed6）
\n– direction: \u5bf9\u80a1\u4ef7\u7684\u53ef\u80fd\u5f71\u54cd（positive\/negative\/neutral）
\n– impact_level: \u5f71\u54cd\u7a0b\u5ea6（high\/medium\/low）
\n– summary: \u4e00\u53e5\u8bdd\u603b\u7ed3\u4e8b\u4ef6（\u4e0d\u8d85\u8fc730\u5b57）<\/p>\n
\u65b0\u95fb\u5185\u5bb9：
\n{news_text}<\/p>\n
\u8bf7\u4ec5\u8fd4\u56deJSON，\u4e0d\u8981\u6709\u5176\u4ed6\u6587\u5b57\u3002"""<\/p>\n
def __init__(self, api_key=None, base_url=None):
\n """\u521d\u59cb\u5316：\u914d\u7f6e\u5927\u6a21\u578bAPI（\u4f7f\u7528OpenAI\u517c\u5bb9\u63a5\u53e3）"""
\n self.api_key = api_key or os.getenv("OPENAI_API_KEY")
\n self.base_url = base_url or os.getenv("OPENAI_BASE_URL")<\/p>\n
def detect(self, news_text: str) -> dict:
\n """
\n \u8bc6\u522b\u5355\u6761\u65b0\u95fb\u662f\u5426\u5305\u542b\u50ac\u5316\u5242\u4e8b\u4ef6<\/p>\n
\u5728\u5b9e\u9645\u90e8\u7f72\u4e2d，\u8fd9\u4e2a\u51fd\u6570\u4f1a\u88ab\u6279\u91cf\u8c03\u7528，
\n \u5904\u7406\u6bcf\u5929\u65b0\u51fa\u73b0\u7684\u6570\u767e\u6761\u516c\u544a\u548c\u65b0\u95fb
\n """
\n try:
\n # \u4f7f\u7528openai\u5e93（pip install openai）
\n from openai import OpenAI
\n client = OpenAI(api_key=self.api_key, base_url=self.base_url)<\/p>\n
response = client.chat.completions.create(
\n model="gpt-4o-mini", # \u4f7f\u7528\u8f7b\u91cf\u6a21\u578b\u964d\u4f4eAPI\u6210\u672c
\n messages=[
\n {"role": "system", "content": "\u4f60\u662f\u4e00\u4e2aA\u80a1\u4e8b\u4ef6\u5206\u6790\u52a9\u624b，\u8bf7\u4e25\u683c\u6309JSON\u683c\u5f0f\u56de\u590d\u3002"},
\n {"role": "user", "content": self.DETECTION_PROMPT.format(
\n news_text=news_text
\n )}
\n ],
\n temperature=0.1, # \u4f4e\u6e29\u5ea6\u4fdd\u8bc1\u8f93\u51fa\u683c\u5f0f\u7a33\u5b9a
\n max_tokens=200
\n )<\/p>\n
result = json.loads(response.choices[0].message.content)
\n return result<\/p>\n
except Exception as e:
\n print(f"\u50ac\u5316\u5242\u8bc6\u522b\u5931\u8d25：{e}")
\n return {"is_catalyst": False, "error": str(e)}<\/p>\n
def batch_detect(self, news_list: list) -> list:
\n """\u6279\u91cf\u8bc6\u522b，\u8fd4\u56de\u6240\u6709\u88ab\u6807\u8bb0\u4e3a\u50ac\u5316\u5242\u7684\u4e8b\u4ef6"""
\n catalysts = []
\n for i, news in enumerate(news_list):
\n result = self.detect(news)
\n if result.get('is_catalyst'):
\n catalysts.append({
\n **result,
\n 'original_news': news,
\n 'detected_at': datetime.now().isoformat()
\n })
\n if (i + 1) % 50 == 0:
\n print(f"\u5df2\u5904\u7406 {i+1}\/{len(news_list)} \u6761\u65b0\u95fb…")
\n print(f"\u5171\u8bc6\u522b\u51fa {len(catalysts)} \u6761\u50ac\u5316\u5242\u4e8b\u4ef6")
\n return catalysts<\/p>\n
# === \u4f7f\u7528\u793a\u4f8b ===
\n# detector = CatalystDetector(api_key="your-api-key")
\n# test_news = "\u8d35\u5dde\u8305\u53f02024\u5e74\u7b2c\u4e09\u5b63\u5ea6\u51c0\u5229\u6da6\u540c\u6bd4\u589e\u957f25%，\u8d85\u51fa\u5e02\u573a\u9884\u671f\u768418%"
\n# result = detector.detect(test_news)
\n# print(json.dumps(result, ensure_ascii=False, indent=2))
\nprint("\u50ac\u5316\u5242\u8bc6\u522b\u6a21\u5757\u52a0\u8f7d\u5b8c\u6210\u3002\u5b9e\u9645\u4f7f\u7528\u9700\u914d\u7f6e\u5927\u6a21\u578bAPI\u5bc6\u94a5\u3002")<\/p>\n
\u524d\u9762\u4e09\u8282\u5206\u522b\u91c7\u96c6\u4e86\u884c\u60c5\u6570\u636e\u3001\u8d22\u52a1\u6570\u636e\u548c\u4e8b\u4ef6\u6570\u636e\u3002\u73b0\u5728\u9700\u8981\u628a\u5b83\u4eec\u7edf\u4e00\u5b58\u50a8\u5230\u4e00\u4e2a\u7ed3\u6784\u5316\u7684\u672c\u5730\u6570\u636e\u5e93\u4e2d\u3002\u63a8\u8350\u4f7f\u7528\u4e24\u79cd\u683c\u5f0f\u7684\u7ec4\u5408：<\/p>\n
| HDF5<\/td>\n | \u5927\u89c4\u6a21\u884c\u60c5\u6570\u636e（\u6570\u5343\u53ea\u80a1\u7968 \u00d7 \u5341\u5e74\u65e5\u7ebf）<\/td>\n | \u8bfb\u5199\u6781\u5feb，\u652f\u6301\u538b\u7f29，\u5355\u6587\u4ef6\u5b58\u50a8<\/td>\n | \u5e76\u53d1\u5199\u5165\u80fd\u529b\u5f31<\/td>\n<\/tr>\n |
| SQLite<\/td>\n | \u8d22\u52a1\u6570\u636e\u3001\u4e8b\u4ef6\u6570\u636e（\u5b57\u6bb5\u7ed3\u6784\u56fa\u5b9a）<\/td>\n | \u652f\u6301SQL\u67e5\u8be2，\u591a\u8868\u5173\u8054<\/td>\n | \u5199\u5165\u901f\u5ea6\u4e0d\u5982HDF5<\/td>\n<\/tr>\n |
| Parquet<\/td>\n | \u4e2d\u95f4\u8ba1\u7b97\u7ed3\u679c\u3001\u56e0\u5b50\u503c<\/td>\n | \u5217\u5f0f\u5b58\u50a8，\u538b\u7f29\u7387\u9ad8<\/td>\n | \u4e0d\u652f\u6301\u539f\u5730\u66f4\u65b0<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n\u4ee3\u7801\u6f14\u793a<\/h4>\n\u4e0b\u9762\u6784\u5efa\u4e00\u4e2a\u7edf\u4e00\u7684\u672c\u5730\u6570\u636e\u8bbf\u95ee\u5c42：<\/p>\n import sqlite3 class LocalDataWarehouse: def __init__(self, base_dir=".\/local_data"): def _init_db(self): # \u4e8b\u4ef6\u8868：\u5b58\u50a8\u8bc6\u522b\u51fa\u7684\u50ac\u5316\u5242\u4e8b\u4ef6 # \u4ea4\u6613\u65e5\u5386\u8868：\u7528\u4e8e\u5feb\u901f\u5224\u65ad\u67d0\u5929\u662f\u5426\u4e3a\u4ea4\u6613\u65e5 conn.commit() def insert_event(self, stock_code: str, event_data: dict): def query_events(self, stock_code=None, days=7): def get_daily_data(self, stock_code): \u5173\u952e\u8981\u70b9<\/h4>\n\n \u5b9e\u6218\u603b\u7ed3<\/h3>\n\u672c\u8282\u5c0f\u6674（\u60c5\u62a5\u5b98）\u5b8c\u6210\u4e86\u4ed6\u7684\u7b2c\u4e00\u6b21"\u5165\u804c\u4efb\u52a1"\u2014\u2014\u642d\u5efa\u4ece\u6570\u636e\u91c7\u96c6\u5230\u672c\u5730\u5b58\u50a8\u7684\u5b8c\u6574\u7ba1\u9053：<\/p>\n \u5e38\u89c1\u8bef\u533a\u63d0\u9192：<\/p>\n
|