克隆策略

Transformer在量化选股中的应用研究

Transformer:Attention is all you need

paper: https://arxiv.org/abs/1706.03762

The naive transformer implemented here for financial time series prediction follows the paper "Attention is all you need":

Given the input (N, T, F),

  1. An embedding layer that maps the input (N, T, F) to representation (N, T, F’);
  2. A positional encoding layer that adds the positional sigmoid;
  3. An encoder that consists of several encoding layers, each of which uses a self-attention layer as the computing module (function of query, key, and value).
  4. A decoder that consists of an MLP (or a Linear layer) that maps the representation of the last time (N, 1, F') into output (N, 1).

    {"description":"实验创建于2017/8/26","graph":{"edges":[{"to_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-15:instruments","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-8:data"},{"to_node_id":"-106:instruments","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-8:data"},{"to_node_id":"-38256:input_1","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-15:data"},{"to_node_id":"-106:features","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"to_node_id":"-113:features","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"to_node_id":"-122:features","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"to_node_id":"-129:features","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"to_node_id":"-251:features","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"to_node_id":"-266:input_2","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"to_node_id":"-288:features","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"to_node_id":"-293:features","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"to_node_id":"-298:input_2","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"to_node_id":"-25911:features","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"to_node_id":"-25911:input_data","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-53:data"},{"to_node_id":"-122:instruments","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-62:data"},{"to_node_id":"-141:instruments","from_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-62:data"},{"to_node_id":"-113:input_data","from_node_id":"-106:data"},{"to_node_id":"-266:input_1","from_node_id":"-113:data"},{"to_node_id":"-129:input_data","from_node_id":"-122:data"},{"to_node_id":"-298:input_1","from_node_id":"-129:data"},{"to_node_id":"-2431:input_2","from_node_id":"-129:data"},{"to_node_id":"-436:input_2","from_node_id":"-251:data"},{"to_node_id":"-2431:input_1","from_node_id":"-436:data_1"},{"to_node_id":"-288:input_data","from_node_id":"-266:data"},{"to_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-53:data2","from_node_id":"-288:data"},{"to_node_id":"-251:input_data","from_node_id":"-293:data"},{"to_node_id":"-293:input_data","from_node_id":"-298:data"},{"to_node_id":"-141:options_data","from_node_id":"-2431:data_1"},{"to_node_id":"-436:input_1","from_node_id":"-25911:data"},{"to_node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-53:data1","from_node_id":"-38256:data"}],"nodes":[{"node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-8","module_id":"BigQuantSpace.instruments.instruments-v2","parameters":[{"name":"start_date","value":"2010-01-01","type":"Literal","bound_global_parameter":null},{"name":"end_date","value":"2017-12-31","type":"Literal","bound_global_parameter":null},{"name":"market","value":"CN_STOCK_A","type":"Literal","bound_global_parameter":null},{"name":"instrument_list","value":"","type":"Literal","bound_global_parameter":null},{"name":"max_count","value":"0","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"rolling_conf","node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-8"}],"output_ports":[{"name":"data","node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-8"}],"cacheable":true,"seq_num":22,"comment":"","comment_collapsed":true},{"node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-15","module_id":"BigQuantSpace.advanced_auto_labeler.advanced_auto_labeler-v2","parameters":[{"name":"label_expr","value":"# #号开始的表示注释\n# 0. 每行一个,顺序执行,从第二个开始,可以使用label字段\n# 1. 可用数据字段见 https://bigquant.com/docs/data_history_data.html\n# 添加benchmark_前缀,可使用对应的benchmark数据\n# 2. 可用操作符和函数见 `表达式引擎 <https://bigquant.com/docs/big_expr.html>`_\n\n# 计算收益:5日收盘价(作为卖出价格)除以明日开盘价(作为买入价格)\nshift(close, -5) / shift(open, -1)-1\n\n# 极值处理:用1%和99%分位的值做clip\nclip(label, all_quantile(label, 0.01), all_quantile(label, 0.99))\n\n# 过滤掉一字涨停的情况 (设置label为NaN,在后续处理和训练中会忽略NaN的label)\nwhere(shift(high, -1) == shift(low, -1), NaN, label)\n","type":"Literal","bound_global_parameter":null},{"name":"start_date","value":"","type":"Literal","bound_global_parameter":null},{"name":"end_date","value":"","type":"Literal","bound_global_parameter":null},{"name":"benchmark","value":"000300.SHA","type":"Literal","bound_global_parameter":null},{"name":"drop_na_label","value":"True","type":"Literal","bound_global_parameter":null},{"name":"cast_label_int","value":"False","type":"Literal","bound_global_parameter":null},{"name":"user_functions","value":"","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"instruments","node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-15"}],"output_ports":[{"name":"data","node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-15"}],"cacheable":true,"seq_num":23,"comment":"","comment_collapsed":true},{"node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24","module_id":"BigQuantSpace.input_features.input_features-v1","parameters":[{"name":"features","value":"close_0\nopen_0\nhigh_0\nlow_0 \namount_0\nturn_0 \nreturn_0\n \nclose_1\nopen_1\nhigh_1\nlow_1\nreturn_1\namount_1\nturn_1\n \nclose_2\nopen_2\nhigh_2\nlow_2\namount_2\nturn_2\nreturn_2\n \nclose_3\nopen_3\nhigh_3\nlow_3\namount_3\nturn_3\nreturn_3\n \nclose_4\nopen_4\nhigh_4\nlow_4\namount_4\nturn_4\nreturn_4\n \nmean(close_0, 5)\nmean(low_0, 5)\nmean(open_0, 5)\nmean(high_0, 5)\nmean(turn_0, 5)\nmean(amount_0, 5)\nmean(return_0, 5)\n \nts_max(close_0, 5)\nts_max(low_0, 5)\nts_max(open_0, 5)\nts_max(high_0, 5)\nts_max(turn_0, 5)\nts_max(amount_0, 5)\nts_max(return_0, 5)\n \nts_min(close_0, 5)\nts_min(low_0, 5)\nts_min(open_0, 5)\nts_min(high_0, 5)\nts_min(turn_0, 5)\nts_min(amount_0, 5)\nts_min(return_0, 5) \n \nstd(close_0, 5)\nstd(low_0, 5)\nstd(open_0, 5)\nstd(high_0, 5)\nstd(turn_0, 5)\nstd(amount_0, 5)\nstd(return_0, 5)\n \nts_rank(close_0, 5)\nts_rank(low_0, 5)\nts_rank(open_0, 5)\nts_rank(high_0, 5)\nts_rank(turn_0, 5)\nts_rank(amount_0, 5)\nts_rank(return_0, 5)\n \ndecay_linear(close_0, 5)\ndecay_linear(low_0, 5)\ndecay_linear(open_0, 5)\ndecay_linear(high_0, 5)\ndecay_linear(turn_0, 5)\ndecay_linear(amount_0, 5)\ndecay_linear(return_0, 5)\n \ncorrelation(volume_0, return_0, 5)\ncorrelation(volume_0, high_0, 5)\ncorrelation(volume_0, low_0, 5)\ncorrelation(volume_0, close_0, 5)\ncorrelation(volume_0, open_0, 5)\ncorrelation(volume_0, turn_0, 5)\n \ncorrelation(return_0, high_0, 5)\ncorrelation(return_0, low_0, 5)\ncorrelation(return_0, close_0, 5)\ncorrelation(return_0, open_0, 5)\ncorrelation(return_0, turn_0, 5)\n \ncorrelation(high_0, low_0, 5)\ncorrelation(high_0, close_0, 5)\ncorrelation(high_0, open_0, 5)\ncorrelation(high_0, turn_0, 5)\n \ncorrelation(low_0, close_0, 5)\ncorrelation(low_0, open_0, 5)\ncorrelation(low_0, turn_0, 5)\n \ncorrelation(close_0, open_0, 5)\ncorrelation(close_0, turn_0, 5)\n\ncorrelation(open_0, turn_0, 5)","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"features_ds","node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24"}],"output_ports":[{"name":"data","node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24"}],"cacheable":true,"seq_num":24,"comment":"","comment_collapsed":true},{"node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-53","module_id":"BigQuantSpace.join.join-v3","parameters":[{"name":"on","value":"date,instrument","type":"Literal","bound_global_parameter":null},{"name":"how","value":"inner","type":"Literal","bound_global_parameter":null},{"name":"sort","value":"False","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"data1","node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-53"},{"name":"data2","node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-53"}],"output_ports":[{"name":"data","node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-53"}],"cacheable":true,"seq_num":25,"comment":"","comment_collapsed":true},{"node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-62","module_id":"BigQuantSpace.instruments.instruments-v2","parameters":[{"name":"start_date","value":"2018-01-01","type":"Literal","bound_global_parameter":null},{"name":"end_date","value":"2021-07-01","type":"Literal","bound_global_parameter":null},{"name":"market","value":"CN_STOCK_A","type":"Literal","bound_global_parameter":null},{"name":"instrument_list","value":"","type":"Literal","bound_global_parameter":null},{"name":"max_count","value":"0","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"rolling_conf","node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-62"}],"output_ports":[{"name":"data","node_id":"287d2cb0-f53c-4101-bdf8-104b137c8601-62"}],"cacheable":true,"seq_num":26,"comment":"预测数据,用于回测和模拟","comment_collapsed":true},{"node_id":"-106","module_id":"BigQuantSpace.general_feature_extractor.general_feature_extractor-v7","parameters":[{"name":"start_date","value":"","type":"Literal","bound_global_parameter":null},{"name":"end_date","value":"","type":"Literal","bound_global_parameter":null},{"name":"before_start_days","value":"10","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"instruments","node_id":"-106"},{"name":"features","node_id":"-106"}],"output_ports":[{"name":"data","node_id":"-106"}],"cacheable":true,"seq_num":27,"comment":"","comment_collapsed":true},{"node_id":"-113","module_id":"BigQuantSpace.derived_feature_extractor.derived_feature_extractor-v3","parameters":[{"name":"date_col","value":"date","type":"Literal","bound_global_parameter":null},{"name":"instrument_col","value":"instrument","type":"Literal","bound_global_parameter":null},{"name":"drop_na","value":"True","type":"Literal","bound_global_parameter":null},{"name":"remove_extra_columns","value":"False","type":"Literal","bound_global_parameter":null},{"name":"user_functions","value":"","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"input_data","node_id":"-113"},{"name":"features","node_id":"-113"}],"output_ports":[{"name":"data","node_id":"-113"}],"cacheable":true,"seq_num":28,"comment":"","comment_collapsed":true},{"node_id":"-122","module_id":"BigQuantSpace.general_feature_extractor.general_feature_extractor-v7","parameters":[{"name":"start_date","value":"","type":"Literal","bound_global_parameter":null},{"name":"end_date","value":"","type":"Literal","bound_global_parameter":null},{"name":"before_start_days","value":"10","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"instruments","node_id":"-122"},{"name":"features","node_id":"-122"}],"output_ports":[{"name":"data","node_id":"-122"}],"cacheable":true,"seq_num":29,"comment":"","comment_collapsed":true},{"node_id":"-129","module_id":"BigQuantSpace.derived_feature_extractor.derived_feature_extractor-v3","parameters":[{"name":"date_col","value":"date","type":"Literal","bound_global_parameter":null},{"name":"instrument_col","value":"instrument","type":"Literal","bound_global_parameter":null},{"name":"drop_na","value":"True","type":"Literal","bound_global_parameter":null},{"name":"remove_extra_columns","value":"False","type":"Literal","bound_global_parameter":null},{"name":"user_functions","value":"","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"input_data","node_id":"-129"},{"name":"features","node_id":"-129"}],"output_ports":[{"name":"data","node_id":"-129"}],"cacheable":true,"seq_num":30,"comment":"","comment_collapsed":true},{"node_id":"-251","module_id":"BigQuantSpace.dl_convert_to_bin.dl_convert_to_bin-v2","parameters":[{"name":"window_size","value":"5","type":"Literal","bound_global_parameter":null},{"name":"feature_clip","value":"3","type":"Literal","bound_global_parameter":null},{"name":"flatten","value":"False","type":"Literal","bound_global_parameter":null},{"name":"window_along_col","value":"instrument","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"input_data","node_id":"-251"},{"name":"features","node_id":"-251"}],"output_ports":[{"name":"data","node_id":"-251"}],"cacheable":true,"seq_num":32,"comment":"","comment_collapsed":true},{"node_id":"-436","module_id":"BigQuantSpace.cached.cached-v3","parameters":[{"name":"run","value":"# Python 代码入口函数,input_1/2/3 对应三个输入端,data_1/2/3 对应三个输出端\ndef bigquant_run(input_1, input_2, input_3):\n # 示例代码如下。在这里编写您的代码\n from sklearn.model_selection import train_test_split\n # train data\n train_data = input_1.read()\n x_train, x_val, y_train, y_val = train_test_split(train_data[\"x\"], train_data['y'], shuffle=True, random_state=2021)\n # val data\n test_data = input_2.read()\n x_test = test_data[\"x\"]\n \n from bigmodels.models.transformer import Transformer\n \n model = Transformer(input_dim=98, embed_dim=256, nhead=8, num_layers=6, dropout=0.1)\n model.compile(device=\"cuda:0\")\n model.fit(x_train, y_train, val_data=(x_val, y_val), batch_size=2048, epochs=10, verbose=1, num_workers=2)\n \n # model.fit(train_data[\"x\"], train_data['y'], batch_size=1024, epochs=2, verbose=1, num_workers=2)\n output = model.predict(x_test)\n \n data_1 = DataSource.write_pickle(output)\n return Outputs(data_1=data_1, data_2=None, data_3=None)","type":"Literal","bound_global_parameter":null},{"name":"post_run","value":"# 后处理函数,可选。输入是主函数的输出,可以在这里对数据做处理,或者返回更友好的outputs数据格式。此函数输出不会被缓存。\ndef bigquant_run(outputs):\n return outputs\n","type":"Literal","bound_global_parameter":null},{"name":"input_ports","value":"","type":"Literal","bound_global_parameter":null},{"name":"params","value":"{}","type":"Literal","bound_global_parameter":null},{"name":"output_ports","value":"","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"input_1","node_id":"-436"},{"name":"input_2","node_id":"-436"},{"name":"input_3","node_id":"-436"}],"output_ports":[{"name":"data_1","node_id":"-436"},{"name":"data_2","node_id":"-436"},{"name":"data_3","node_id":"-436"}],"cacheable":false,"seq_num":33,"comment":"Transformer训练和预测","comment_collapsed":false},{"node_id":"-266","module_id":"BigQuantSpace.standardlize.standardlize-v8","parameters":[{"name":"columns_input","value":"[]","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"input_1","node_id":"-266"},{"name":"input_2","node_id":"-266"}],"output_ports":[{"name":"data","node_id":"-266"}],"cacheable":true,"seq_num":34,"comment":"","comment_collapsed":true},{"node_id":"-288","module_id":"BigQuantSpace.fillnan.fillnan-v1","parameters":[{"name":"fill_value","value":"0.0","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"input_data","node_id":"-288"},{"name":"features","node_id":"-288"}],"output_ports":[{"name":"data","node_id":"-288"}],"cacheable":true,"seq_num":35,"comment":"","comment_collapsed":true},{"node_id":"-293","module_id":"BigQuantSpace.fillnan.fillnan-v1","parameters":[{"name":"fill_value","value":"0.0","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"input_data","node_id":"-293"},{"name":"features","node_id":"-293"}],"output_ports":[{"name":"data","node_id":"-293"}],"cacheable":true,"seq_num":36,"comment":"","comment_collapsed":true},{"node_id":"-298","module_id":"BigQuantSpace.standardlize.standardlize-v8","parameters":[{"name":"columns_input","value":"[]","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"input_1","node_id":"-298"},{"name":"input_2","node_id":"-298"}],"output_ports":[{"name":"data","node_id":"-298"}],"cacheable":true,"seq_num":37,"comment":"","comment_collapsed":true},{"node_id":"-2431","module_id":"BigQuantSpace.cached.cached-v3","parameters":[{"name":"run","value":"# Python 代码入口函数,input_1/2/3 对应三个输入端,data_1/2/3 对应三个输出端\ndef bigquant_run(input_1, input_2, input_3):\n # 示例代码如下。在这里编写您的代码\n pred_label = input_1.read_pickle()\n \n df = input_2.read_df()\n df = pd.DataFrame({'pred_label':pred_label[:], 'instrument':df.instrument, 'date':df.date})\n df.sort_values(['date','pred_label'],inplace=True, ascending=[True,False])\n return Outputs(data_1=DataSource.write_df(df), data_2=None, data_3=None)\n","type":"Literal","bound_global_parameter":null},{"name":"post_run","value":"# 后处理函数,可选。输入是主函数的输出,可以在这里对数据做处理,或者返回更友好的outputs数据格式。此函数输出不会被缓存。\ndef bigquant_run(outputs):\n return outputs\n","type":"Literal","bound_global_parameter":null},{"name":"input_ports","value":"","type":"Literal","bound_global_parameter":null},{"name":"params","value":"{}","type":"Literal","bound_global_parameter":null},{"name":"output_ports","value":"","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"input_1","node_id":"-2431"},{"name":"input_2","node_id":"-2431"},{"name":"input_3","node_id":"-2431"}],"output_ports":[{"name":"data_1","node_id":"-2431"},{"name":"data_2","node_id":"-2431"},{"name":"data_3","node_id":"-2431"}],"cacheable":false,"seq_num":41,"comment":"","comment_collapsed":true},{"node_id":"-141","module_id":"BigQuantSpace.trade.trade-v4","parameters":[{"name":"start_date","value":"","type":"Literal","bound_global_parameter":null},{"name":"end_date","value":"","type":"Literal","bound_global_parameter":null},{"name":"initialize","value":"# 回测引擎:初始化函数,只执行一次\ndef bigquant_run(context):\n # 加载预测数据\n context.ranker_prediction = context.options['data'].read_df()\n\n # 系统已经设置了默认的交易手续费和滑点,要修改手续费可使用如下函数\n context.set_commission(PerOrder(buy_cost=0.001, sell_cost=0.001, min_cost=5))\n # 预测数据,通过options传入进来,使用 read_df 函数,加载到内存 (DataFrame)\n # 设置买入的股票数量,这里买入预测股票列表排名靠前的5只\n stock_count = 20\n # 每只的股票的权重,如下的权重分配会使得靠前的股票分配多一点的资金,[0.339160, 0.213986, 0.169580, ..]\n context.stock_weights = T.norm([1 / math.log(i + 2) for i in range(0, stock_count)])\n # 设置每只股票占用的最大资金比例\n context.max_cash_per_instrument = 0.2\n context.options['hold_days'] = 5\n","type":"Literal","bound_global_parameter":null},{"name":"handle_data","value":"# 回测引擎:每日数据处理函数,每天执行一次\ndef bigquant_run(context, data):\n # 按日期过滤得到今日的预测数据\n ranker_prediction = context.ranker_prediction[\n context.ranker_prediction.date == data.current_dt.strftime('%Y-%m-%d')]\n\n # 1. 资金分配\n # 平均持仓时间是hold_days,每日都将买入股票,每日预期使用 1/hold_days 的资金\n # 实际操作中,会存在一定的买入误差,所以在前hold_days天,等量使用资金;之后,尽量使用剩余资金(这里设置最多用等量的1.5倍)\n is_staging = context.trading_day_index < context.options['hold_days'] # 是否在建仓期间(前 hold_days 天)\n cash_avg = context.portfolio.portfolio_value / context.options['hold_days']\n cash_for_buy = min(context.portfolio.cash, (1 if is_staging else 1.5) * cash_avg)\n cash_for_sell = cash_avg - (context.portfolio.cash - cash_for_buy)\n positions = {e.symbol: p.amount * p.last_sale_price\n for e, p in context.perf_tracker.position_tracker.positions.items()}\n\n # 2. 生成卖出订单:hold_days天之后才开始卖出;对持仓的股票,按机器学习算法预测的排序末位淘汰\n if not is_staging and cash_for_sell > 0:\n equities = {e.symbol: e for e, p in context.perf_tracker.position_tracker.positions.items()}\n instruments = list(reversed(list(ranker_prediction.instrument[ranker_prediction.instrument.apply(\n lambda x: x in equities and not context.has_unfinished_sell_order(equities[x]))])))\n # print('rank order for sell %s' % instruments)\n for instrument in instruments:\n context.order_target(context.symbol(instrument), 0)\n cash_for_sell -= positions[instrument]\n if cash_for_sell <= 0:\n break\n\n # 3. 生成买入订单:按机器学习算法预测的排序,买入前面的stock_count只股票\n buy_cash_weights = context.stock_weights\n buy_instruments = list(ranker_prediction.instrument[:len(buy_cash_weights)])\n max_cash_per_instrument = context.portfolio.portfolio_value * context.max_cash_per_instrument\n for i, instrument in enumerate(buy_instruments):\n cash = cash_for_buy * buy_cash_weights[i]\n if cash > max_cash_per_instrument - positions.get(instrument, 0):\n # 确保股票持仓量不会超过每次股票最大的占用资金量\n cash = max_cash_per_instrument - positions.get(instrument, 0)\n if cash > 0:\n context.order_value(context.symbol(instrument), cash)\n","type":"Literal","bound_global_parameter":null},{"name":"prepare","value":"# 回测引擎:准备数据,只执行一次\ndef bigquant_run(context):\n pass\n","type":"Literal","bound_global_parameter":null},{"name":"before_trading_start","value":"","type":"Literal","bound_global_parameter":null},{"name":"volume_limit","value":0.025,"type":"Literal","bound_global_parameter":null},{"name":"order_price_field_buy","value":"open","type":"Literal","bound_global_parameter":null},{"name":"order_price_field_sell","value":"close","type":"Literal","bound_global_parameter":null},{"name":"capital_base","value":1000000,"type":"Literal","bound_global_parameter":null},{"name":"auto_cancel_non_tradable_orders","value":"True","type":"Literal","bound_global_parameter":null},{"name":"data_frequency","value":"daily","type":"Literal","bound_global_parameter":null},{"name":"price_type","value":"后复权","type":"Literal","bound_global_parameter":null},{"name":"product_type","value":"股票","type":"Literal","bound_global_parameter":null},{"name":"plot_charts","value":"True","type":"Literal","bound_global_parameter":null},{"name":"backtest_only","value":"False","type":"Literal","bound_global_parameter":null},{"name":"benchmark","value":"000300.SHA","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"instruments","node_id":"-141"},{"name":"options_data","node_id":"-141"},{"name":"history_ds","node_id":"-141"},{"name":"benchmark_ds","node_id":"-141"},{"name":"trading_calendar","node_id":"-141"}],"output_ports":[{"name":"raw_perf","node_id":"-141"}],"cacheable":false,"seq_num":42,"comment":"","comment_collapsed":true},{"node_id":"-25911","module_id":"BigQuantSpace.dl_convert_to_bin.dl_convert_to_bin-v2","parameters":[{"name":"window_size","value":"5","type":"Literal","bound_global_parameter":null},{"name":"feature_clip","value":"3","type":"Literal","bound_global_parameter":null},{"name":"flatten","value":"False","type":"Literal","bound_global_parameter":null},{"name":"window_along_col","value":"instrument","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"input_data","node_id":"-25911"},{"name":"features","node_id":"-25911"}],"output_ports":[{"name":"data","node_id":"-25911"}],"cacheable":true,"seq_num":2,"comment":"","comment_collapsed":true},{"node_id":"-38256","module_id":"BigQuantSpace.standardlize.standardlize-v8","parameters":[{"name":"columns_input","value":"label","type":"Literal","bound_global_parameter":null}],"input_ports":[{"name":"input_1","node_id":"-38256"},{"name":"input_2","node_id":"-38256"}],"output_ports":[{"name":"data","node_id":"-38256"}],"cacheable":true,"seq_num":1,"comment":"","comment_collapsed":true}],"node_layout":"<node_postions><node_position Node='287d2cb0-f53c-4101-bdf8-104b137c8601-8' Position='324,-13,200,200'/><node_position Node='287d2cb0-f53c-4101-bdf8-104b137c8601-15' Position='21,167,200,200'/><node_position Node='287d2cb0-f53c-4101-bdf8-104b137c8601-24' Position='775,-126,200,200'/><node_position Node='287d2cb0-f53c-4101-bdf8-104b137c8601-53' Position='275,445,200,200'/><node_position Node='287d2cb0-f53c-4101-bdf8-104b137c8601-62' Position='1132.300048828125,18.4692325592041,200,200'/><node_position Node='-106' Position='436,108,200,200'/><node_position Node='-113' Position='430,172,200,200'/><node_position Node='-122' Position='1135,144,200,200'/><node_position Node='-129' Position='1135,237,200,200'/><node_position Node='-251' Position='1121,558,200,200'/><node_position Node='-436' Position='558,693,200,200'/><node_position Node='-266' Position='436,251,200,200'/><node_position Node='-288' Position='433,319,200,200'/><node_position Node='-293' Position='1127,452,200,200'/><node_position Node='-298' Position='1130,337,200,200'/><node_position Node='-2431' Position='562,844,200,200'/><node_position Node='-141' Position='488,1014,200,200'/><node_position Node='-25911' Position='281,561,200,200'/><node_position Node='-38256' Position='9,260,200,200'/></node_postions>"},"nodes_readonly":false,"studio_version":"v2"}
    In [1]:
    # 本代码由可视化策略环境自动生成 2021年10月26日 16:27
    # 本代码单元只能在可视化模式下编辑。您也可以拷贝代码,粘贴到新建的代码单元或者策略,然后修改。
    
    
    # Python 代码入口函数,input_1/2/3 对应三个输入端,data_1/2/3 对应三个输出端
    def m33_run_bigquant_run(input_1, input_2, input_3):
        # 示例代码如下。在这里编写您的代码
        from sklearn.model_selection import train_test_split
        # train data
        train_data = input_1.read()
        x_train, x_val, y_train, y_val = train_test_split(train_data["x"], train_data['y'], shuffle=True, random_state=2021)
        # val data
        test_data = input_2.read()
        x_test = test_data["x"]
        
        from bigmodels.models.transformer import Transformer
        
        model = Transformer(input_dim=98, embed_dim=256, nhead=8, num_layers=6, dropout=0.1)
        model.compile(device="cuda:0")
        model.fit(x_train, y_train, val_data=(x_val, y_val), batch_size=2048, epochs=10, verbose=1, num_workers=2)
        
        # model.fit(train_data["x"], train_data['y'], batch_size=1024, epochs=2, verbose=1, num_workers=2)
        output = model.predict(x_test)
        
        data_1 = DataSource.write_pickle(output)
        return Outputs(data_1=data_1, data_2=None, data_3=None)
    # 后处理函数,可选。输入是主函数的输出,可以在这里对数据做处理,或者返回更友好的outputs数据格式。此函数输出不会被缓存。
    def m33_post_run_bigquant_run(outputs):
        return outputs
    
    # Python 代码入口函数,input_1/2/3 对应三个输入端,data_1/2/3 对应三个输出端
    def m41_run_bigquant_run(input_1, input_2, input_3):
        # 示例代码如下。在这里编写您的代码
        pred_label = input_1.read_pickle()
        
        df = input_2.read_df()
        df = pd.DataFrame({'pred_label':pred_label[:], 'instrument':df.instrument, 'date':df.date})
        df.sort_values(['date','pred_label'],inplace=True, ascending=[True,False])
        return Outputs(data_1=DataSource.write_df(df), data_2=None, data_3=None)
    
    # 后处理函数,可选。输入是主函数的输出,可以在这里对数据做处理,或者返回更友好的outputs数据格式。此函数输出不会被缓存。
    def m41_post_run_bigquant_run(outputs):
        return outputs
    
    # 回测引擎:初始化函数,只执行一次
    def m42_initialize_bigquant_run(context):
        # 加载预测数据
        context.ranker_prediction = context.options['data'].read_df()
    
        # 系统已经设置了默认的交易手续费和滑点,要修改手续费可使用如下函数
        context.set_commission(PerOrder(buy_cost=0.001, sell_cost=0.001, min_cost=5))
        # 预测数据,通过options传入进来,使用 read_df 函数,加载到内存 (DataFrame)
        # 设置买入的股票数量,这里买入预测股票列表排名靠前的5只
        stock_count = 20
        # 每只的股票的权重,如下的权重分配会使得靠前的股票分配多一点的资金,[0.339160, 0.213986, 0.169580, ..]
        context.stock_weights = T.norm([1 / math.log(i + 2) for i in range(0, stock_count)])
        # 设置每只股票占用的最大资金比例
        context.max_cash_per_instrument = 0.2
        context.options['hold_days'] = 5
    
    # 回测引擎:每日数据处理函数,每天执行一次
    def m42_handle_data_bigquant_run(context, data):
        # 按日期过滤得到今日的预测数据
        ranker_prediction = context.ranker_prediction[
            context.ranker_prediction.date == data.current_dt.strftime('%Y-%m-%d')]
    
        # 1. 资金分配
        # 平均持仓时间是hold_days,每日都将买入股票,每日预期使用 1/hold_days 的资金
        # 实际操作中,会存在一定的买入误差,所以在前hold_days天,等量使用资金;之后,尽量使用剩余资金(这里设置最多用等量的1.5倍)
        is_staging = context.trading_day_index < context.options['hold_days'] # 是否在建仓期间(前 hold_days 天)
        cash_avg = context.portfolio.portfolio_value / context.options['hold_days']
        cash_for_buy = min(context.portfolio.cash, (1 if is_staging else 1.5) * cash_avg)
        cash_for_sell = cash_avg - (context.portfolio.cash - cash_for_buy)
        positions = {e.symbol: p.amount * p.last_sale_price
                     for e, p in context.perf_tracker.position_tracker.positions.items()}
    
        # 2. 生成卖出订单:hold_days天之后才开始卖出;对持仓的股票,按机器学习算法预测的排序末位淘汰
        if not is_staging and cash_for_sell > 0:
            equities = {e.symbol: e for e, p in context.perf_tracker.position_tracker.positions.items()}
            instruments = list(reversed(list(ranker_prediction.instrument[ranker_prediction.instrument.apply(
                    lambda x: x in equities and not context.has_unfinished_sell_order(equities[x]))])))
            # print('rank order for sell %s' % instruments)
            for instrument in instruments:
                context.order_target(context.symbol(instrument), 0)
                cash_for_sell -= positions[instrument]
                if cash_for_sell <= 0:
                    break
    
        # 3. 生成买入订单:按机器学习算法预测的排序,买入前面的stock_count只股票
        buy_cash_weights = context.stock_weights
        buy_instruments = list(ranker_prediction.instrument[:len(buy_cash_weights)])
        max_cash_per_instrument = context.portfolio.portfolio_value * context.max_cash_per_instrument
        for i, instrument in enumerate(buy_instruments):
            cash = cash_for_buy * buy_cash_weights[i]
            if cash > max_cash_per_instrument - positions.get(instrument, 0):
                # 确保股票持仓量不会超过每次股票最大的占用资金量
                cash = max_cash_per_instrument - positions.get(instrument, 0)
            if cash > 0:
                context.order_value(context.symbol(instrument), cash)
    
    # 回测引擎:准备数据,只执行一次
    def m42_prepare_bigquant_run(context):
        pass
    
    
    m22 = M.instruments.v2(
        start_date='2010-01-01',
        end_date='2017-12-31',
        market='CN_STOCK_A',
        instrument_list='',
        max_count=0
    )
    
    m23 = M.advanced_auto_labeler.v2(
        instruments=m22.data,
        label_expr="""# #号开始的表示注释
    # 0. 每行一个,顺序执行,从第二个开始,可以使用label字段
    # 1. 可用数据字段见 https://bigquant.com/docs/data_history_data.html
    #   添加benchmark_前缀,可使用对应的benchmark数据
    # 2. 可用操作符和函数见 `表达式引擎 <https://bigquant.com/docs/big_expr.html>`_
    
    # 计算收益:5日收盘价(作为卖出价格)除以明日开盘价(作为买入价格)
    shift(close, -5) / shift(open, -1)-1
    
    # 极值处理:用1%和99%分位的值做clip
    clip(label, all_quantile(label, 0.01), all_quantile(label, 0.99))
    
    # 过滤掉一字涨停的情况 (设置label为NaN,在后续处理和训练中会忽略NaN的label)
    where(shift(high, -1) == shift(low, -1), NaN, label)
    """,
        start_date='',
        end_date='',
        benchmark='000300.SHA',
        drop_na_label=True,
        cast_label_int=False
    )
    
    m1 = M.standardlize.v8(
        input_1=m23.data,
        columns_input='label'
    )
    
    m24 = M.input_features.v1(
        features="""close_0
    open_0
    high_0
    low_0 
    amount_0
    turn_0 
    return_0
     
    close_1
    open_1
    high_1
    low_1
    return_1
    amount_1
    turn_1
     
    close_2
    open_2
    high_2
    low_2
    amount_2
    turn_2
    return_2
     
    close_3
    open_3
    high_3
    low_3
    amount_3
    turn_3
    return_3
     
    close_4
    open_4
    high_4
    low_4
    amount_4
    turn_4
    return_4
     
    mean(close_0, 5)
    mean(low_0, 5)
    mean(open_0, 5)
    mean(high_0, 5)
    mean(turn_0, 5)
    mean(amount_0, 5)
    mean(return_0, 5)
     
    ts_max(close_0, 5)
    ts_max(low_0, 5)
    ts_max(open_0, 5)
    ts_max(high_0, 5)
    ts_max(turn_0, 5)
    ts_max(amount_0, 5)
    ts_max(return_0, 5)
     
    ts_min(close_0, 5)
    ts_min(low_0, 5)
    ts_min(open_0, 5)
    ts_min(high_0, 5)
    ts_min(turn_0, 5)
    ts_min(amount_0, 5)
    ts_min(return_0, 5) 
     
    std(close_0, 5)
    std(low_0, 5)
    std(open_0, 5)
    std(high_0, 5)
    std(turn_0, 5)
    std(amount_0, 5)
    std(return_0, 5)
     
    ts_rank(close_0, 5)
    ts_rank(low_0, 5)
    ts_rank(open_0, 5)
    ts_rank(high_0, 5)
    ts_rank(turn_0, 5)
    ts_rank(amount_0, 5)
    ts_rank(return_0, 5)
     
    decay_linear(close_0, 5)
    decay_linear(low_0, 5)
    decay_linear(open_0, 5)
    decay_linear(high_0, 5)
    decay_linear(turn_0, 5)
    decay_linear(amount_0, 5)
    decay_linear(return_0, 5)
     
    correlation(volume_0, return_0, 5)
    correlation(volume_0, high_0, 5)
    correlation(volume_0, low_0, 5)
    correlation(volume_0, close_0, 5)
    correlation(volume_0, open_0, 5)
    correlation(volume_0, turn_0, 5)
      
    correlation(return_0, high_0, 5)
    correlation(return_0, low_0, 5)
    correlation(return_0, close_0, 5)
    correlation(return_0, open_0, 5)
    correlation(return_0, turn_0, 5)
     
    correlation(high_0, low_0, 5)
    correlation(high_0, close_0, 5)
    correlation(high_0, open_0, 5)
    correlation(high_0, turn_0, 5)
     
    correlation(low_0, close_0, 5)
    correlation(low_0, open_0, 5)
    correlation(low_0, turn_0, 5)
     
    correlation(close_0, open_0, 5)
    correlation(close_0, turn_0, 5)
    
    correlation(open_0, turn_0, 5)"""
    )
    
    m27 = M.general_feature_extractor.v7(
        instruments=m22.data,
        features=m24.data,
        start_date='',
        end_date='',
        before_start_days=10
    )
    
    m28 = M.derived_feature_extractor.v3(
        input_data=m27.data,
        features=m24.data,
        date_col='date',
        instrument_col='instrument',
        drop_na=True,
        remove_extra_columns=False
    )
    
    m34 = M.standardlize.v8(
        input_1=m28.data,
        input_2=m24.data,
        columns_input='[]'
    )
    
    m35 = M.fillnan.v1(
        input_data=m34.data,
        features=m24.data,
        fill_value='0.0'
    )
    
    m25 = M.join.v3(
        data1=m1.data,
        data2=m35.data,
        on='date,instrument',
        how='inner',
        sort=False
    )
    
    m2 = M.dl_convert_to_bin.v2(
        input_data=m25.data,
        features=m24.data,
        window_size=5,
        feature_clip=3,
        flatten=False,
        window_along_col='instrument'
    )
    
    m26 = M.instruments.v2(
        start_date='2018-01-01',
        end_date='2021-07-01',
        market='CN_STOCK_A',
        instrument_list='',
        max_count=0
    )
    
    m29 = M.general_feature_extractor.v7(
        instruments=m26.data,
        features=m24.data,
        start_date='',
        end_date='',
        before_start_days=10
    )
    
    m30 = M.derived_feature_extractor.v3(
        input_data=m29.data,
        features=m24.data,
        date_col='date',
        instrument_col='instrument',
        drop_na=True,
        remove_extra_columns=False
    )
    
    m37 = M.standardlize.v8(
        input_1=m30.data,
        input_2=m24.data,
        columns_input='[]'
    )
    
    m36 = M.fillnan.v1(
        input_data=m37.data,
        features=m24.data,
        fill_value='0.0'
    )
    
    m32 = M.dl_convert_to_bin.v2(
        input_data=m36.data,
        features=m24.data,
        window_size=5,
        feature_clip=3,
        flatten=False,
        window_along_col='instrument'
    )
    
    m33 = M.cached.v3(
        input_1=m2.data,
        input_2=m32.data,
        run=m33_run_bigquant_run,
        post_run=m33_post_run_bigquant_run,
        input_ports='',
        params='{}',
        output_ports='',
        m_cached=False
    )
    
    m41 = M.cached.v3(
        input_1=m33.data_1,
        input_2=m30.data,
        run=m41_run_bigquant_run,
        post_run=m41_post_run_bigquant_run,
        input_ports='',
        params='{}',
        output_ports='',
        m_cached=False
    )
    
    m42 = M.trade.v4(
        instruments=m26.data,
        options_data=m41.data_1,
        start_date='',
        end_date='',
        initialize=m42_initialize_bigquant_run,
        handle_data=m42_handle_data_bigquant_run,
        prepare=m42_prepare_bigquant_run,
        volume_limit=0.025,
        order_price_field_buy='open',
        order_price_field_sell='close',
        capital_base=1000000,
        auto_cancel_non_tradable_orders=True,
        data_frequency='daily',
        price_type='后复权',
        product_type='股票',
        plot_charts=True,
        backtest_only=False,
        benchmark='000300.SHA'
    )
    
    epoch 0   |  train_loss 1.03101|  vall_loss 0.98711|  0:03:21s
    epoch 1   |  train_loss 0.98289|  vall_loss 0.98238|  0:06:42s
    epoch 2   |  train_loss 0.98027|  vall_loss 0.98351|  0:10:03s
    epoch 3   |  train_loss 0.97960|  vall_loss 0.98329|  0:13:24s
    epoch 4   |  train_loss 0.97926|  vall_loss 0.98756|  0:16:46s
    epoch 5   |  train_loss 0.97805|  vall_loss 0.98104|  0:20:07s
    epoch 6   |  train_loss 0.97538|  vall_loss 0.97769|  0:23:29s
    epoch 7   |  train_loss 0.97130|  vall_loss 0.97642|  0:26:49s
    epoch 8   |  train_loss 0.96649|  vall_loss 0.97246|  0:30:14s
    epoch 9   |  train_loss 0.96351|  vall_loss 0.97224|  0:33:37s
    best loss: 0.972243 @ 9
    
    • 收益率260.14%
    • 年化收益率46.28%
    • 基准收益率29.74%
    • 阿尔法0.39
    • 贝塔0.96
    • 夏普比率1.29
    • 胜率0.52
    • 盈亏比1.24
    • 收益波动率30.8%
    • 信息比率0.09
    • 最大回撤25.71%
    bigcharts-data-start/{"__type":"tabs","__id":"bigchart-3455396ed4994e9184105b811eff7cda"}/bigcharts-data-end