{"id":472,"date":"2024-07-01T12:02:03","date_gmt":"2024-07-01T04:02:03","guid":{"rendered":"http:\/\/123.60.176.65\/?p=472"},"modified":"2024-07-01T12:02:03","modified_gmt":"2024-07-01T04:02:03","slug":"ai%e9%a1%be%e9%97%ae%e6%8a%95%e8%b5%84%e4%b9%8b%e9%ab%98%e7%ba%a7%e7%ad%96%e7%95%a5%e4%b8%80%ef%bc%9a%e5%b7%b4%e8%8f%b2%e7%89%b9%e7%9a%84alpha%ef%bc%88%e5%bb%ba%e8%ae%ae%e6%94%b6%e8%97%8f%ef%bc%89","status":"publish","type":"post","link":"https:\/\/laoyulaoyu.com\/index.php\/2024\/07\/01\/ai%e9%a1%be%e9%97%ae%e6%8a%95%e8%b5%84%e4%b9%8b%e9%ab%98%e7%ba%a7%e7%ad%96%e7%95%a5%e4%b8%80%ef%bc%9a%e5%b7%b4%e8%8f%b2%e7%89%b9%e7%9a%84alpha%ef%bc%88%e5%bb%ba%e8%ae%ae%e6%94%b6%e8%97%8f%ef%bc%89\/","title":{"rendered":"AI\u987e\u6295\u9ad8\u7ea7\u7b56\u7565\u4e4b\u4e00\uff1a\u5df4\u83f2\u7279\u7684alpha\uff08\u5efa\u8bae\u6536\u85cf\uff09"},"content":{"rendered":"<div class=\"wp-block-image is-style-default\">\n<figure class=\"aligncenter size-large\"><img decoding=\"async\" src=\"http:\/\/123.60.176.65\/wp-content\/uploads\/2024\/06\/bft-1024x547.webp\" alt=\"\" class=\"wp-image-562\"\/><\/figure>\n<\/div>\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<pre class=\"wp-block-verse\"><strong>\u5199\u5728\u524d\u9762\u7684\u8bdd\uff1a<\/strong>\u5728\u8fd9\u4e2a<strong>\u5408\u96c6<\/strong>\u7684\u5f00\u5934\uff0c\u6211\u4eec\u6765\u5199\u5199<mark style=\"background-color:rgba(0, 0, 0, 0)\" class=\"has-inline-color has-vivid-cyan-blue-color\">\u5df4\u83f2\u7279\u7684alpha<\/mark>\u3002\u5df4\u83f2\u7279\u7684\u6295\u8d44\u7b56\u7565\u548c\u5176\u6240\u5b9e\u73b0\u7684\u8d85\u989d\u56de\u62a5\uff08\u5373Alpha\uff09\u4e00\u76f4\u662f\u91d1\u878d\u5b66\u754c\u548c\u5b9e\u52a1\u754c\u7684\u7814\u7a76\u7126\u70b9\u3002Alpha\u662f\u4e00\u4e2a\u8861\u91cf\u6295\u8d44\u6027\u80fd\u7684\u6307\u6807\uff0c\u4ee3\u8868\u4e86\u4e00\u4e2a\u6295\u8d44\u7ec4\u5408\u76f8\u5bf9\u4e8e\u5176\u57fa\u51c6\u6307\u6570\u7684\u8d85\u989d\u56de\u62a5\u3002\u5728\u6c83\u4f26\u00b7\u5df4\u83f2\u7279\u7684\u6848\u4f8b\u4e2d\uff0c\u8fd9\u610f\u5473\u7740\u4ed6\u7ba1\u7406\u7684\u4f2f\u514b\u5e0c\u5c14\u00b7\u54c8\u6492\u97e6\u516c\u53f8\u7684\u6295\u8d44<mark style=\"background-color:rgba(0, 0, 0, 0)\" class=\"has-inline-color has-black-color\">\u56de\u62a5\u8d85\u8fc7\u4e86\u4e00\u822c\u5e02\u573a\u7684\u8868\u73b0\u3002<\/mark><\/pre>\n<\/blockquote>\n\n\n\n<p>\u7ecf\u6d4e\u5b66\u5bb6\u548c\u91d1\u878d\u5206\u6790\u5e08\u7ecf\u5e38\u7814\u7a76\u5df4\u83f2\u7279\u7684\u6295\u8d44\u7b56\u7565\uff0c\u8bd5\u56fe\u91cf\u5316\u4ed6\u7684Alpha\u5e76\u7406\u89e3\u5176\u80fd\u591f\u6301\u7eed\u6218\u80dc\u5e02\u573a\u7684\u539f\u56e0\u3002\u7814\u7a76\u663e\u793a\uff0c\u5df4\u83f2\u7279\u7684Alpha\u90e8\u5206\u6765\u81ea\u4e8e\u4ed6\u9009\u80a1\u7684\u80fd\u529b\u548c\u884c\u4e1a\u914d\u7f6e\u7684\u7b56\u7565\uff0c\u800c\u4e14\u4ed6\u90a3\u770b\u4f3c\u7b80\u5355\u7684\u4ef7\u503c\u6295\u8d44\u7b56\u7565\u80cc\u540e\u5b9e\u9645\u4e0a\u662f\u590d\u6742\u548c\u7cbe\u7ec6\u7684\u8d22\u52a1\u5206\u6790\u53ca\u5e02\u573a\u6d1e\u5bdf\u3002<\/p>\n\n\n\n<p>\u5df4\u83f2\u7279\u7684\u6210\u529f\u548c\u9ad8Alpha\u4e5f\u5e38\u88ab\u89c6\u4e3a\u6295\u8d44\u9886\u57df\u4e2d\u7684\u4e00\u4e2a\u6807\u6746\uff0c\u6fc0\u52b1\u4e86\u65e0\u6570\u6295\u8d44\u8005\u548c\u57fa\u91d1\u7ecf\u7406\u53bb\u6a21\u4eff\u4ed6\u7684\u7b56\u7565\u548c\u539f\u5219\u3002\u4e0d\u8fc7\uff0c\u5df4\u83f2\u7279\u672c\u4eba\u5e38\u5e38<strong>\u5f3a\u8c03\uff0c\u8010\u5fc3\u3001\u7eaa\u5f8b\u548c\u6b63\u786e\u7684\u5fc3\u6001<\/strong>\u662f\u6295\u8d44\u6210\u529f\u7684\u5173\u952e\u3002<\/p>\n\n\n\n<p>\u5728\u300a\u5df4\u83f2\u7279\u7684alpha\u300b\u6587\u7ae0\u91cc\uff0c\u540e\u4eba\u628a\u5df4\u83f2\u7279\u7684\u6536\u76ca\u5206\u6210\u516d\u4e2a\u7ef4\u5ea6\uff0c\u5206\u522b\u662f<strong>\u5e02\u573a\uff0c\u4f30\u503c\uff0c\u89c4\u6a21\uff0c\u52a8\u91cf\uff0c\u8d28\u91cf\u548c\u6ce2\u52a8\u7387\u516d\u4e2a\u7ef4\u5ea6<\/strong>\u3002\u6211\u4eec\u4eca\u5929\u5c31\u5f00\u59cb\u590d\u73b0\u5176\u4e2d\u7684\u539f\u7406\uff0c\u9664\u53bb\u5e02\u573a\u7ef4\u5ea6\uff0c\u6211\u4eec\u4ece\u5176\u4ed6\u4e94\u4e2a\u7ef4\u5ea6\u5206\u522b\u6311\u9009\u56e0\u5b50\uff0c\u603b\u51716\u4e2a\uff0c\u7ec4\u62106\u56e0\u5b50\u6a21\u578b\u3002<\/p>\n\n\n\n<p>\u5bf9\u4e8e\u56e0\u5b50\u7684\u5904\u7406\uff0c\u7531\u4e8e\u56e0\u5b50\u6765\u81ea\u4e0d\u540c\u7ef4\u5ea6\uff0c\u6240\u4ee5\u65e0\u9700\u8fdb\u884c\u964d\u7ef4\u6216\u8005\u56e0\u5b50\u6b63\u4ea4\u5904\u7406\u6765\u89e3\u51b3\u5b83\u7684\u76f8\u5173\u6027\u95ee\u9898\uff0c\u6240\u4ee5\u7b80\u5355\u8fdb\u884c\u4e86\u53bb\u6781\u503c\u548c\u6807\u51c6\u5316\u5904\u7406\u3002<\/p>\n\n\n\n<p>\u5bf9\u4e8e\u80a1\u7968\u5217\u8868\uff0c\u5df2\u7ecf\u8fdb\u884c\u5254\u9664ST\uff0c\u4e0a\u5e02\u672a\u6ee160\u5929\u7684\u65b0\u80a1\uff0c\u505c\u724c\u80a1\u548c\u5f00\u76d8\u6da8\u505c\u80a1\u3002<\/p>\n\n\n\n<p>\u5bf9\u4e8e\u6253\u5206\u65b9\u5f0f\uff1a\u9488\u5bf9\u5347\u5e8f\u56e0\u5b50\u4e58\u4ee5-1\uff1b\u9488\u5bf9\u964d\u5e8f\u56e0\u5b50\u4e58\u4ee51\uff0c\u6700\u540e\u8fdb\u884c\u53e0\u52a0\u3002<\/p>\n\n\n\n<p>\u5728\u62e9\u65f6\u65b9\u9762\uff0c\u91c7\u7528RSRS\u7684\u65b9\u5f0f\u5bf9\u6307\u6570\u8fdb\u884c\u62e9\u65f6\u4fe1\u53f7\u3002<\/p>\n\n\n\n<p>\u8d44\u91d1\u89c4\u6a21\u572810000000\uff0c20\u65e5\u8c03\u4ed3\uff0c\u56de\u6d4b\u65f6\u95f4\u4ece2010-01-01\u52302018-11-08&nbsp;<\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><strong><strong>\u7b97\u6cd5\u4ee3\u7801\u5982\u4e0b\uff1a<\/strong><\/strong><\/h2>\n\n\n\n<pre class=\"wp-block-code\"><code><code>\"\"\"\n\u4ee5\u4e0b\u4ee3\u7801\u8fd0\u884c\u540e\u6536\u76ca\u8fd8\u884c\uff0c\u53ef\u4ee5\u4f7f\u7528\u3002\n\n\"\"\"\nimport pandas as pd\nimport numpy as np\nimport datetime as dt\nimport talib as ta\nfrom datetime import date, timedelta\nimport statsmodels.api as sm\n\n# \u521d\u59cb\u5316\u8d26\u6237\ndef init(context):\n    set_params(context)\n    set_variables(context)\n    set_backtest()\n    run_daily(stop_loss)\n\n# \u8bbe\u7f6e\u7b56\u7565\u53c2\u6570\ndef set_params(context):\n    g.tc = 20  # \u8c03\u4ed3\u9891\u7387\n    g.t = 0\n    g.big_small = 'big'  # big\u662f\u964d\u5e8f\uff0csmall\u4e3a\u5347\u5e8f\n    context.stock = '000300.SH'\n    g.long_pct = 0.05\n    g.stock = '000300.SH'  # \u62e9\u65f6\u9009\u53d6\u7684\u6307\u6570\n    g.total_positionprevious = 0  # \u4ed3\u4f4d\n    g.N = 18  # RSRS\u9009\u53d6\u7684\u56de\u5f52\u957f\u5ea6\n    g.M = 1100  # RSRS\u5747\u503c\u7a97\u53e3\n\n# \u8bbe\u7f6e\u53d8\u91cf\ndef set_variables(context):\n    context.X_length = 11\n    context.flag = True\n    g.buy = 0.7  # \u4e70\u5165\u9600\u95e8\n    g.sell = -0.7  # \u5356\u51fa\u9600\u95e8\n    g.ans = &#91;]\n    g.ans_rightdev = &#91;]\n\n# \u8bbe\u7f6e\u56de\u6d4b\ndef set_backtest():\n    set_benchmark('000300.SH')  # \u8bbe\u7f6e\u57fa\u51c6\n    set_slippage(PriceSlippage(0.002))  # \u8bbe\u7f6e\u53ef\u53d8\u6ed1\u70b9\n\n# \u4e2a\u80a1\u6b62\u635f\ndef stop_loss(context, bar_dict):\n    for stock in list(context.portfolio.positions):\n        cumulative_return = bar_dict&#91;stock].close \/ context.portfolio.positions&#91;stock].cost_basis\n        if cumulative_return &lt; 0.9:\n            order_target_value(stock, 0)\n\n# \u5904\u7406K\u7ebf\ndef handle_bar(context, bar_dict):\n    stock = g.stock\n    beta = 0\n    r2 = 0\n    prices = history(stock, &#91;'high', 'low'], g.N, '1d', False, 'pre', is_panel=1)\n    highs = prices.high\n    lows = prices.low\n    X = sm.add_constant(lows)\n    model = sm.OLS(highs, X)\n    # \u5f97\u5230beta\n    beta = model.fit().params&#91;1]\n    # \u5c06\u65b0\u7684beta\u6dfb\u52a0\u5230\u88c5\u6709\u5386\u53f2\u6570\u636e\u5217\u8868\n    g.ans.append(beta)\n    # \u5f97\u5230rsquare\u6570\u636e\n    r2 = model.fit().rsquared\n    # \u5c06\u65b0\u7684rsquare\u6dfb\u52a0\u5230\u88c5\u6709\u5386\u53f2\u6570\u636e\u5217\u8868\n    g.ans_rightdev.append(r2)\n    # \u4e3a\u4e86\u6807\u51c6\u5316\u5f53\u4e0b\u7684beta\u6570\u503c\uff0c\u62ff\u8fc7\u53bb1100\u5929\u7684\u6570\u636e\u4f5c\u4e3a\u5747\u503c\u7684\u7a97\u53e3\n    section = g.ans&#91;-g.M:]\n    # \u8ba1\u7b97\u5747\u503c\u5e8f\u5217\n    mu = np.mean(section)\n    # \u8ba1\u7b97\u6807\u51c6\u5316RSRS\u6307\u6807\u5e8f\u5217\n    sigma = np.std(section)\n    zscore = (section&#91;-1] - mu) \/ sigma\n    # \u8ba1\u7b97\u53f3\u504fRSRS\u6807\u51c6\u5206\uff0c\u5c31\u662f\u5c06\u6807\u51c6\u5316\u540e\u7684beta\u6570\u636e\u4e58\u4ee5\u539f\u59cbbeta\u518d\u4e58\u4ee5\u62df\u5408\u5ea6\n    zscore_rightdev = zscore * beta * r2\n    # \u6839\u636e\u4ea4\u6613\u4fe1\u53f7\u4e70\u5165\u5356\u51fa\n    if zscore_rightdev &gt; g.buy:\n        total_position = 1\n    elif zscore_rightdev &lt; g.sell:\n        total_position = 0\n    else:\n        total_position = g.total_positionprevious\n    if (g.total_positionprevious!= total_position) or (g.t % g.tc == 0):\n        g.total_positionprevious = total_position\n    last_date = get_last_datetime().strftime('%Y%m%d')\n    stock_list = list(get_all_securities('stock', date=last_date).index)\n    # \u5bf9stock_list\u8fdb\u884c\u53bb\u9664st\uff0c\u505c\u724c\u7b49\u5904\u7406\n    stock_list = fun_unpaused(bar_dict, stock_list)\n    stock_list = fun_st(bar_dict, stock_list)\n    stock_list = fun_highlimit(bar_dict, stock_list)\n    stock_list = fun_remove_new(stock_list, 60)\n    # \u4ee5\u4e0b\u662f\u5404\u5355\u56e0\u5b50\n    # \u89c4\u6a21\u56e0\u5b50\n    cap_df = market_cap(stock_list, 'valuation_market_cap', last_date)\n    cap_df = cap_df * -1\n    # \u4f30\u503c\u56e0\u5b50\n    PB_df = PB(stock_list, 'valuation_pb', last_date)\n    PB_df = PB_df * -1\n    # \u52a8\u91cf\u56e0\u5b50\n    MTM20_df = MTM20(stock_list, 'MTM20')\n    MTM20_df = MTM20_df * -1\n    # \u8d28\u91cf\u56e0\u5b50\n    # 1.ROE\uff08\u9ad8\u5229\u6da6\uff09\n    roe_df = roe(stock_list, 'profit_roe_ths', last_date)\n    # 2.\u51c0\u5229\u6da6\u540c\u6bd4\u589e\u957f\u7387\uff08\u9ad8\u6210\u957f\uff09\n    net_profit_growth_ratio_df = net_profit_growth_ratio(stock_list, 'growth_net_profit_growth_ratio', last_date)\n    # \u6ce2\u52a8\u7387\u56e0\u5b50\n    ATR20_df = ATR20(stock_list, 'ATR20')\n    ATR20_df = ATR20_df * -1\n    # \u5408\u5e76\u591a\u56e0\u5b50\n    concat_obj = &#91;cap_df, PB_df, MTM20_df, roe_df, net_profit_growth_ratio_df, ATR20_df]\n    df = pd.concat(concat_obj, axis=1)\n    df = df.dropna()\n    # log.info(type(df))\n    sum = df.sum(axis=1)\n    # log.info(sum)\n    # \u8fdb\u884c\u6392\u5e8f\n    if g.big_small == 'big':\n        # \u6309\u7167\u5927\u6392\u5e8f\n        sum.sort_values(ascending=False, inplace=True)\n    if g.big_small =='small':\n        # \u6309\u7167\u5c0f\u6392\u5e8f\n        sum.sort_values(ascending=True, inplace=True)\n    # \u6839\u636e\u6bd4\u4f8b\u53d6\u51fa\u6392\u5e8f\u540e\u9760\u524d\u90e8\u5206\n    stock_list1 = sum&#91;0:int(len(stock_list) * g.long_pct)].index\n    # log.info(stock_list1)\n    buy_list = &#91;]\n    for stock in stock_list1:\n        buy_list.append(stock)\n    # \u4e70\u5356\u64cd\u4f5c\n    for stock in list(context.portfolio.positions):\n        if stock not in buy_list:\n            order_target(stock, 0)\n    cash = context.portfolio.portfolio_value\n    position = cash * g.total_positionprevious\n    num = int(len(stock_list) * g.long_pct)\n    ## \u4e70\u5165\n    for stock in buy_list:\n        order_target_value(stock, position \/ num)\n    g.t = g.t + 1\n\n# \u4ee5\u4e0b\u662f\u5355\u56e0\u5b50\ndef market_cap(stocklist, factor, last_date):\n    # \u53d6\u6570\u636e\n    df = get_fundamentals(query(valuation.symbol, valuation.market_cap).filter(valuation.symbol.in_(stocklist)),\n                          date=last_date)\n    # log.info(df)\n    df = df.set_index('valuation_symbol')\n    # \u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\n    after_MAD = MAD(factor, df)\n    # z-score\u6cd5\u6807\u51c6\u5316\n    after_zscore = zscore(factor, after_MAD)\n    return after_zscore\n\ndef PB(stocklist, factor, last_date):\n    # \u53d6\u6570\u636e\n    df = get_fundamentals(query(valuation.symbol, valuation.pb).filter(valuation.symbol.in_(stocklist)),\n                          date=last_date)\n    df = df.set_index('valuation_symbol')\n    # \u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\n    after_MAD = MAD(factor, df)\n    # z-score\u6cd5\u6807\u51c6\u5316\n    after_zscore = zscore(factor, after_MAD)\n    return after_zscore\n\ndef MTM20(stocklist, factor):\n    # \u53d6\u6570\u636e\n    for stock in stocklist:\n        df1 = history(stock, &#91;'close'], 20, '1d')\n        # log.info(df1)\n        s = pd.DataFrame(&#91;(df1&#91;'close']&#91;-1] - df1&#91;'close']&#91;0]) \/ df1&#91;'close']&#91;0]], index=&#91;stock])\n        # log.info(s)\n        if 'df' in locals():\n            df = df.append(s)\n        else:\n            df = s\n    # log.info(df)\n    df.columns = &#91;'MTM20']\n    df.index.name = 'valuation_symbol'\n    # \u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\n    after_MAD = MAD(factor, df)\n    # z-score\u6cd5\u6807\u51c6\u5316\n    after_zscore = zscore(factor, after_MAD)\n    return after_zscore\n\ndef roe(stocklist, factor, last_date):\n    # \u53d6\u6570\u636e\n    df = get_fundamentals(query(valuation.symbol, profit.roe_ths).filter(valuation.symbol.in_(stocklist)),\n                          date=last_date)\nlog.info(df)\ndf = df.set_index('valuation_symbol')\n\u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\nafter_MAD = MAD(factor, df)\nz-score\u6cd5\u6807\u51c6\u5316\nafter_zscore = zscore(factor, after_MAD)\nreturn after_zscore\ndef net_profit_growth_ratio(stocklist, factor, last_date):\n\u53d6\u6570\u636e\ndf = get_fundamentals(query(valuation.symbol, growth.net_profit_growth_ratio).filter(valuation.symbol.in_(stocklist)),\ndate=last_date)\nlog.info(df)\ndf = df.set_index('valuation_symbol')\n\u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\nafter_MAD = MAD(factor, df)\nz-score\u6cd5\u6807\u51c6\u5316\nafter_zscore = zscore(factor, after_MAD)\nreturn after_zscore\ndef ATR20(stocklist, new_factor):\n\u53d6\u6570\u636e\nfor stock in stocklist:\nData_ATR = history(stock, &#91;'close', 'high', 'low'], 20, '1d')\nclose_ATR = np.array(Data_ATR&#91;'close'])\nhigh_ATR = np.array(Data_ATR&#91;'high'])\nlow_ATR = np.array(Data_ATR&#91;'low'])\n'''\nif np.isnan(close_ATR).any():\ncontinue\n'''\nATR = ta.ATR(high_ATR, low_ATR, close_ATR, timeperiod=1)\nindices = ~np.isnan(ATR)\nresult = np.average(ATR&#91;indices])\ns = pd.Series(result.astype(float), index=&#91;stock])\nif 'ATR_df' in locals():\nATR_df = ATR_df.append(s)\nelse:\nATR_df = s\ndf = ATR_df.to_frame()\ndf.index.name = 'valuation_symbol'\ndf.columns = &#91;new_factor]\n\u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\nafter_MAD = MAD(new_factor, df)\nz-score\u6cd5\u6807\u51c6\u5316\nafter_zscore = zscore(new_factor, after_MAD)\nreturn after_zscore\n\u4ee5\u4e0b\u662f\u8fdb\u884c\u56e0\u5b50\u6570\u636e\u5904\u7406\uff0c\u5bf9\u56e0\u5b50\u8fdb\u884cMAD\u53bb\u6781\u503c\uff0c\u4ee5\u53ca\u6807\u51c6\u5316\u5904\u7406\ndef MAD(factor, df):\n\u53d6\u5f97\u4e2d\u4f4d\u6570\nmedian = df&#91;factor].median()\n\u53d6\u5f97\u6570\u636e\u4e0e\u4e2d\u4f4d\u6570\u5dee\u503c\ndf1 = df - median\n\u53d6\u5f97\u5dee\u503c\u7edd\u5bf9\u503c\ndf1 = df1.abs()\n\u53d6\u5f97\u7edd\u5bf9\u4e2d\u4f4d\u6570\nMAD = df1&#91;factor].median()\n\u5f97\u5230\u6570\u636e\u4e0a\u4e0b\u8fb9\u754c\nextreme_upper = median + 3 * 1.483 * MAD\nextreme_lower = median - 3 * 1.483 * MAD\n\u5c06\u6570\u636e\u4e0a\u4e0b\u8fb9\u754c\u5916\u7684\u6570\u503c\u5f52\u5230\u8fb9\u754c\u4e0a\ndf.ix&#91;(df&#91;factor] &lt; extreme_lower), factor] = extreme_lower\ndf.ix&#91;(df&#91;factor] &gt; extreme_upper), factor] = extreme_upper\nreturn df\nz-score\u6807\u51c6\u5316\ndef zscore(factor, df):\n\u53d6\u5f97\u5747\u503c\nmean = df&#91;factor].mean()\n\u53d6\u5f97\u6807\u51c6\u5dee\nstd = df&#91;factor].std()\n\u53d6\u5f97\u6807\u51c6\u5316\u540e\u6570\u636e\ndf = (df - mean) \/ std\nreturn df\n\u4ee5\u4e0b\u5bf9\u80a1\u7968\u5217\u8868\u8fdb\u884c\u53bb\u9664ST\uff0c\u505c\u724c\uff0c\u53bb\u65b0\u80a1\uff0c\u4ee5\u53ca\u53bb\u9664\u5f00\u76d8\u6da8\u505c\u80a1\n\u53bb\u9664\u5f00\u76d8\u6da8\u505c\u80a1\u7968\ndef fun_highlimit(bar_dict, stock_list):\nreturn &#91;stock for stock in stock_list if bar_dict&#91;stock].open!= bar_dict&#91;stock].high_limit]\n\u53bb\u9664st\u80a1\u7968\ndef fun_st(bar_dict, stock_list):\nreturn &#91;stock for stock in stock_list if not bar_dict&#91;stock].is_st]\ndef fun_unpaused(bar_dict, stock_list):\nreturn &#91;s for s in stock_list if not bar_dict&#91;s].is_paused]\ndef fun_remove_new(_stock_list, days):\ndeltaDate = get_datetime() - dt.timedelta(days)\nstock_list = &#91;]\nfor stock in _stock_list:\nif get_security_info(stock).listed_date &lt; deltaDate:\nstock_list.append(stock)\nreturn stock_list\nimport pandas as pd\nimport numpy as np\nimport datetime as dt\nimport talib as ta\nfrom datetime import date, timedelta\nimport statsmodels.api as sm\n\n# \u521d\u59cb\u5316\u8d26\u6237\ndef init(context):\n    set_params(context)\n    set_variables(context)\n    set_backtest()\n    run_daily(stop_loss)\n\n# \u8bbe\u7f6e\u7b56\u7565\u53c2\u6570\ndef set_params(context):\n    g.tc = 20  # \u8c03\u4ed3\u9891\u7387\n    g.t = 0\n    g.big_small = 'big'  # big\u662f\u964d\u5e8f\uff0csmall\u4e3a\u5347\u5e8f\n    context.stock = '000300.SH'\n    g.long_pct = 0.05\n    g.stock = '000300.SH'  # \u62e9\u65f6\u9009\u53d6\u7684\u6307\u6570\n    g.total_positionprevious = 0  # \u4ed3\u4f4d\n    g.N = 18  # RSRS\u9009\u53d6\u7684\u56de\u5f52\u957f\u5ea6\n    g.M = 1100  # RSRS\u5747\u503c\u7a97\u53e3\n\n# \u8bbe\u7f6e\u53d8\u91cf\ndef set_variables(context):\n    context.X_length = 11\n    context.flag = True\n    g.buy = 0.7  # \u4e70\u5165\u9600\u95e8\n    g.sell = -0.7  # \u5356\u51fa\u9600\u95e8\n    g.ans = &#91;]\n    g.ans_rightdev = &#91;]\n\n# \u8bbe\u7f6e\u56de\u6d4b\ndef set_backtest():\n    set_benchmark('000300.SH')  # \u8bbe\u7f6e\u57fa\u51c6\n    set_slippage(PriceSlippage(0.002))  # \u8bbe\u7f6e\u53ef\u53d8\u6ed1\u70b9\n\n# \u4e2a\u80a1\u6b62\u635f\ndef stop_loss(context, bar_dict):\n    for stock in list(context.portfolio.positions):\n        cumulative_return = bar_dict&#91;stock].close \/ context.portfolio.positions&#91;stock].cost_basis\n        if cumulative_return &lt; 0.9:\n            order_target_value(stock, 0)\n\n# \u5904\u7406K\u7ebf\ndef handle_bar(context, bar_dict):\n    stock = g.stock\n    beta = 0\n    r2 = 0\n    prices = history(stock, &#91;'high', 'low'], g.N, '1d', False, 'pre', is_panel=1)\n    highs = prices.high\n    lows = prices.low\n    X = sm.add_constant(lows)\n    model = sm.OLS(highs, X)\n    # \u5f97\u5230beta\n    beta = model.fit().params&#91;1]\n    # \u5c06\u65b0\u7684beta\u6dfb\u52a0\u5230\u88c5\u6709\u5386\u53f2\u6570\u636e\u5217\u8868\n    g.ans.append(beta)\n    # \u5f97\u5230rsquare\u6570\u636e\n    r2 = model.fit().rsquared\n    # \u5c06\u65b0\u7684rsquare\u6dfb\u52a0\u5230\u88c5\u6709\u5386\u53f2\u6570\u636e\u5217\u8868\n    g.ans_rightdev.append(r2)\n    # \u4e3a\u4e86\u6807\u51c6\u5316\u5f53\u4e0b\u7684beta\u6570\u503c\uff0c\u62ff\u8fc7\u53bb1100\u5929\u7684\u6570\u636e\u4f5c\u4e3a\u5747\u503c\u7684\u7a97\u53e3\n    section = g.ans&#91;-g.M:]\n    # \u8ba1\u7b97\u5747\u503c\u5e8f\u5217\n    mu = np.mean(section)\n    # \u8ba1\u7b97\u6807\u51c6\u5316RSRS\u6307\u6807\u5e8f\u5217\n    sigma = np.std(section)\n    zscore = (section&#91;-1] - mu) \/ sigma\n    # \u8ba1\u7b97\u53f3\u504fRSRS\u6807\u51c6\u5206\uff0c\u5c31\u662f\u5c06\u6807\u51c6\u5316\u540e\u7684beta\u6570\u636e\u4e58\u4ee5\u539f\u59cbbeta\u518d\u4e58\u4ee5\u62df\u5408\u5ea6\n    zscore_rightdev = zscore * beta * r2\n    # \u6839\u636e\u4ea4\u6613\u4fe1\u53f7\u4e70\u5165\u5356\u51fa\n    if zscore_rightdev &gt; g.buy:\n        total_position = 1\n    elif zscore_rightdev &lt; g.sell:\n        total_position = 0\n    else:\n        total_position = g.total_positionprevious\n    if (g.total_positionprevious!= total_position) or (g.t % g.tc == 0):\n        g.total_positionprevious = total_position\n    last_date = get_last_datetime().strftime('%Y%m%d')\n    stock_list = list(get_all_securities('stock', date=last_date).index)\n    # \u5bf9stock_list\u8fdb\u884c\u53bb\u9664st\uff0c\u505c\u724c\u7b49\u5904\u7406\n    stock_list = fun_unpaused(bar_dict, stock_list)\n    stock_list = fun_st(bar_dict, stock_list)\n    stock_list = fun_highlimit(bar_dict, stock_list)\n    stock_list = fun_remove_new(stock_list, 60)\n    # \u4ee5\u4e0b\u662f\u5404\u5355\u56e0\u5b50\n    # \u89c4\u6a21\u56e0\u5b50\n    cap_df = market_cap(stock_list, 'valuation_market_cap', last_date)\n    cap_df = cap_df * -1\n    # \u4f30\u503c\u56e0\u5b50\n    PB_df = PB(stock_list, 'valuation_pb', last_date)\n    PB_df = PB_df * -1\n    # \u52a8\u91cf\u56e0\u5b50\n    MTM20_df = MTM20(stock_list, 'MTM20')\n    MTM20_df = MTM20_df * -1\n    # \u8d28\u91cf\u56e0\u5b50\n    # 1.ROE\uff08\u9ad8\u5229\u6da6\uff09\n    roe_df = roe(stock_list, 'profit_roe_ths', last_date)\n    # 2.\u51c0\u5229\u6da6\u540c\u6bd4\u589e\u957f\u7387\uff08\u9ad8\u6210\u957f\uff09\n    net_profit_growth_ratio_df = net_profit_growth_ratio(stock_list, 'growth_net_profit_growth_ratio', last_date)\n    # \u6ce2\u52a8\u7387\u56e0\u5b50\n    ATR20_df = ATR20(stock_list, 'ATR20')\n    ATR20_df = ATR20_df * -1\n    # \u5408\u5e76\u591a\u56e0\u5b50\n    concat_obj = &#91;cap_df, PB_df, MTM20_df, roe_df, net_profit_growth_ratio_df, ATR20_df]\n    df = pd.concat(concat_obj, axis=1)\n    df = df.dropna()\n    # log.info(type(df))\n    sum = df.sum(axis=1)\n    # log.info(sum)\n    # \u8fdb\u884c\u6392\u5e8f\n    if g.big_small == 'big':\n        # \u6309\u7167\u5927\u6392\u5e8f\n        sum.sort_values(ascending=False, inplace=True)\n    if g.big_small =='small':\n        # \u6309\u7167\u5c0f\u6392\u5e8f\n        sum.sort_values(ascending=True, inplace=True)\n    # \u6839\u636e\u6bd4\u4f8b\u53d6\u51fa\u6392\u5e8f\u540e\u9760\u524d\u90e8\u5206\n    stock_list1 = sum&#91;0:int(len(stock_list) * g.long_pct)].index\n    # log.info(stock_list1)\n    buy_list = &#91;]\n    for stock in stock_list1:\n        buy_list.append(stock)\n    # \u4e70\u5356\u64cd\u4f5c\n    for stock in list(context.portfolio.positions):\n        if stock not in buy_list:\n            order_target(stock, 0)\n    cash = context.portfolio.portfolio_value\n    position = cash * g.total_positionprevious\n    num = int(len(stock_list) * g.long_pct)\n    ## \u4e70\u5165\n    for stock in buy_list:\n        order_target_value(stock, position \/ num)\n    g.t = g.t + 1\n\n# \u4ee5\u4e0b\u662f\u5355\u56e0\u5b50\ndef market_cap(stocklist, factor, last_date):\n    # \u53d6\u6570\u636e\n    df = get_fundamentals(query(valuation.symbol, valuation.market_cap).filter(valuation.symbol.in_(stocklist)),\n                          date=last_date)\n    # log.info(df)\n    df = df.set_index('valuation_symbol')\n    # \u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\n    after_MAD = MAD(factor, df)\n    # z-score\u6cd5\u6807\u51c6\u5316\n    after_zscore = zscore(factor, after_MAD)\n    return after_zscore\n\ndef PB(stocklist, factor, last_date):\n    # \u53d6\u6570\u636e\n    df = get_fundamentals(query(valuation.symbol, valuation.pb).filter(valuation.symbol.in_(stocklist)),\n                          date=last_date)\n    df = df.set_index('valuation_symbol')\n    # \u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\n    after_MAD = MAD(factor, df)\n    # z-score\u6cd5\u6807\u51c6\u5316\n    after_zscore = zscore(factor, after_MAD)\n    return after_zscore\n\ndef MTM20(stocklist, factor):\n    # \u53d6\u6570\u636e\n    for stock in stocklist:\n        df1 = history(stock, &#91;'close'], 20, '1d')\n        # log.info(df1)\n        s = pd.DataFrame(&#91;(df1&#91;'close']&#91;-1] - df1&#91;'close']&#91;0]) \/ df1&#91;'close']&#91;0]], index=&#91;stock])\n        # log.info(s)\n        if 'df' in locals():\n            df = df.append(s)\n        else:\n            df = s\n    # log.info(df)\n    df.columns = &#91;'MTM20']\n    df.index.name = 'valuation_symbol'\n    # \u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\n    after_MAD = MAD(factor, df)\n    # z-score\u6cd5\u6807\u51c6\u5316\n    after_zscore = zscore(factor, after_MAD)\n    return after_zscore\n\ndef roe(stocklist, factor, last_date):\n    # \u53d6\u6570\u636e\n    df = get_fundamentals(query(valuation.symbol, profit.roe_ths).filter(valuation.symbol.in_(stocklist)),\n                          date=last_date)\nlog.info(df)\ndf = df.set_index('valuation_symbol')\n\u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\nafter_MAD = MAD(factor, df)\nz-score\u6cd5\u6807\u51c6\u5316\nafter_zscore = zscore(factor, after_MAD)\nreturn after_zscore\ndef net_profit_growth_ratio(stocklist, factor, last_date):\n\u53d6\u6570\u636e\ndf = get_fundamentals(query(valuation.symbol, growth.net_profit_growth_ratio).filter(valuation.symbol.in_(stocklist)),\ndate=last_date)\nlog.info(df)\ndf = df.set_index('valuation_symbol')\n\u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\nafter_MAD = MAD(factor, df)\nz-score\u6cd5\u6807\u51c6\u5316\nafter_zscore = zscore(factor, after_MAD)\nreturn after_zscore\ndef ATR20(stocklist, new_factor):\n\u53d6\u6570\u636e\nfor stock in stocklist:\nData_ATR = history(stock, &#91;'close', 'high', 'low'], 20, '1d')\nclose_ATR = np.array(Data_ATR&#91;'close'])\nhigh_ATR = np.array(Data_ATR&#91;'high'])\nlow_ATR = np.array(Data_ATR&#91;'low'])\n'''\nif np.isnan(close_ATR).any():\ncontinue\n'''\nATR = ta.ATR(high_ATR, low_ATR, close_ATR, timeperiod=1)\nindices = ~np.isnan(ATR)\nresult = np.average(ATR&#91;indices])\ns = pd.Series(result.astype(float), index=&#91;stock])\nif 'ATR_df' in locals():\nATR_df = ATR_df.append(s)\nelse:\nATR_df = s\ndf = ATR_df.to_frame()\ndf.index.name = 'valuation_symbol'\ndf.columns = &#91;new_factor]\n\u7edd\u5bf9\u4e2d\u4f4d\u6570\u6cd5\u53d6\u6781\u503c\nafter_MAD = MAD(new_factor, df)\nz-score\u6cd5\u6807\u51c6\u5316\nafter_zscore = zscore(new_factor, after_MAD)\nreturn after_zscore\n\u4ee5\u4e0b\u662f\u8fdb\u884c\u56e0\u5b50\u6570\u636e\u5904\u7406\uff0c\u5bf9\u56e0\u5b50\u8fdb\u884cMAD\u53bb\u6781\u503c\uff0c\u4ee5\u53ca\u6807\u51c6\u5316\u5904\u7406\ndef MAD(factor, df):\n\u53d6\u5f97\u4e2d\u4f4d\u6570\nmedian = df&#91;factor].median()\n\u53d6\u5f97\u6570\u636e\u4e0e\u4e2d\u4f4d\u6570\u5dee\u503c\ndf1 = df - median\n\u53d6\u5f97\u5dee\u503c\u7edd\u5bf9\u503c\ndf1 = df1.abs()\n\u53d6\u5f97\u7edd\u5bf9\u4e2d\u4f4d\u6570\nMAD = df1&#91;factor].median()\n\u5f97\u5230\u6570\u636e\u4e0a\u4e0b\u8fb9\u754c\nextreme_upper = median + 3 * 1.483 * MAD\nextreme_lower = median - 3 * 1.483 * MAD\n\u5c06\u6570\u636e\u4e0a\u4e0b\u8fb9\u754c\u5916\u7684\u6570\u503c\u5f52\u5230\u8fb9\u754c\u4e0a\ndf.ix&#91;(df&#91;factor] &lt; extreme_lower), factor] = extreme_lower\ndf.ix&#91;(df&#91;factor] &gt; extreme_upper), factor] = extreme_upper\nreturn df\nz-score\u6807\u51c6\u5316\ndef zscore(factor, df):\n\u53d6\u5f97\u5747\u503c\nmean = df&#91;factor].mean()\n\u53d6\u5f97\u6807\u51c6\u5dee\nstd = df&#91;factor].std()\n\u53d6\u5f97\u6807\u51c6\u5316\u540e\u6570\u636e\ndf = (df - mean) \/ std\nreturn df\n\u4ee5\u4e0b\u5bf9\u80a1\u7968\u5217\u8868\u8fdb\u884c\u53bb\u9664ST\uff0c\u505c\u724c\uff0c\u53bb\u65b0\u80a1\uff0c\u4ee5\u53ca\u53bb\u9664\u5f00\u76d8\u6da8\u505c\u80a1\n\u53bb\u9664\u5f00\u76d8\u6da8\u505c\u80a1\u7968\ndef fun_highlimit(bar_dict, stock_list):\nreturn &#91;stock for stock in stock_list if bar_dict&#91;stock].open!= bar_dict&#91;stock].high_limit]\n\u53bb\u9664st\u80a1\u7968\ndef fun_st(bar_dict, stock_list):\nreturn &#91;stock for stock in stock_list if not bar_dict&#91;stock].is_st]\ndef fun_unpaused(bar_dict, stock_list):\nreturn &#91;s for s in stock_list if not bar_dict&#91;s].is_paused]\ndef fun_remove_new(_stock_list, days):\ndeltaDate = get_datetime() - dt.timedelta(days)\nstock_list = &#91;]\nfor stock in _stock_list:\nif get_security_info(stock).listed_date &lt; deltaDate:\nstock_list.append(stock)\nreturn stock_list<\/code><\/code><\/pre>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"has-text-align-center has-cyan-bluish-gray-color has-text-color has-link-color wp-elements-bd402b7c146898dcf3d05749282e5860\"><strong>\u672c\u6587\u5185\u5bb9\u4ec5\u4ec5\u662f\u6280\u672f\u63a2\u8ba8\u548c\u5b66\u4e60\uff0c\u5e76\u4e0d\u6784\u6210\u4efb\u4f55\u6295\u8d44\u5efa\u8bae\u3002<\/strong><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u5199\u5728\u524d\u9762\u7684\u8bdd\uff1a\u5728\u8fd9\u4e2a\u5408\u96c6\u7684\u5f00\u5934\uff0c\u6211\u4eec\u6765\u5199\u5199\u5df4\u83f2\u7279\u7684&#8230;<\/p>\n<div class=\"more-link-wrapper\"><a class=\"more-link\" href=\"https:\/\/laoyulaoyu.com\/index.php\/2024\/07\/01\/ai%e9%a1%be%e9%97%ae%e6%8a%95%e8%b5%84%e4%b9%8b%e9%ab%98%e7%ba%a7%e7%ad%96%e7%95%a5%e4%b8%80%ef%bc%9a%e5%b7%b4%e8%8f%b2%e7%89%b9%e7%9a%84alpha%ef%bc%88%e5%bb%ba%e8%ae%ae%e6%94%b6%e8%97%8f%ef%bc%89\/\">Continue reading<span class=\"screen-reader-text\">AI\u987e\u6295\u9ad8\u7ea7\u7b56\u7565\u4e4b\u4e00\uff1a\u5df4\u83f2\u7279\u7684alpha\uff08\u5efa\u8bae\u6536\u85cf\uff09<\/span><\/a><\/div>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[4],"class_list":["post-472","post","type-post","status-publish","format-standard","hentry","category-aiinvest","tag-ai","entry"],"_links":{"self":[{"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/posts\/472","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/comments?post=472"}],"version-history":[{"count":0,"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/posts\/472\/revisions"}],"wp:attachment":[{"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/media?parent=472"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/categories?post=472"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/tags?post=472"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}