【宽客学院】滚动训练模块使用简介

用户成长系列
标签: #<Tag:0x00007f4ce8cc45e8>

(iQuant) #1

通过滚动训练模块可以实现训练集的定期更新轮换,本文我们会简单介绍如何使用这个模块。

一、什么是滚动训练

为了尽量避免策略失效,我们可以定期更新训练集数据,也就是通过滚动训练的方式更新预测模型以适应最新市场行情的变化,例如:

我们希望训练一个模型,训练集数据从2016-01-01起到最新数据止,每次模型训练所用数据最少train_data_min_days天,最大不超过train_data_max_days天,每train_update_days天更新一下模型。

我们生成N次滚动训练,对应训练集数据时间范围是:

a. 第一次滚动训练的开始日期为设置的start_date, 第一次滚动训练的结束日期为start_date + train_data_min_days对应的日期;

b. 第二次滚动训练的结束日期为:第一次滚动训练的结束日期+模型更新天数*1, 第二次滚动训练的训练集数据起始日期按下述逻辑确定:

  • 如果 第二次滚动训练的结束日期 - 最大数据天数 >= 设置的start_date, 则第二次滚动训练的训练集数据起始日期 = 第二次滚动训练的结束日期-最大数据天数
  • 如果 第二次滚动训练的结束日期 - 最大数据天数 < 设置的start_date, 则第二次滚动训练的训练集数据起始日期 = 设置的start_date

上述时间段的确定如下图所示

c. 第N次滚动训练的结束日期为:第一次滚动训练的结束日期+模型更新天数*(N-1)。

第N次滚动训练的训练集数据起始日期按下述逻辑确定

  • 如果 第N次滚动训练的结束日期 - 最大数据天数 >= 设置的start_date, 则第N次滚动训练的训练集数据起始日期 = 第N次滚动训练的结束日期-最大数据天数
  • 如果 第N次滚动训练的结束日期 - 最大数据天数 < 设置的start_date, 则第N次滚动训练的训练集数据起始日期 = 设置的start_date

上述生成的N个模型对应的预测集数据时间范围是:第N次滚动训练结束日期 + 1天~第N次滚动训练结束日期 + 模型更新天数

从上述逻辑可以看出:

  • 当train_data_max_days > train_data_min_days + train_update_days 时,前几次滚动训练的训练集每次都从设置的start_date开始逐渐增加长度到train_data_max_days;
  • 当train_data_max_days <= train_data_min_days + train_update_days 时,每次滚动训练的训练集数据长度保持不变,始终为train_data_max_days;

典型的应用场景实例如下:

  • 例1:每250个交易日构建并更新一次模型来预测未来250个交易日可以设置为:
    train_data_max_days=250,train_data_min_days=250,train_update_days=250

  • 例2:每22天用过去250天的数据更新一次模型并预测未来22个交易日可以设置为:
    train_data_max_days=250,train_data_min_days=250,train_update_days=22

二、滚动训练设置步骤

第一步:

新建一个可视化模板策略,拖拽“高级优化”下的滚动训练模块至画布,此时训练集和预测集的各日期控制模块参数设置均失效,由“滚动训练”模块中的参数覆盖并按上述逻辑集中控制,该模块不需要与其它模块连线;

第二步:

根据实际策略中的模块编号对应修改训练集证券代码列表模块id、预测集证券代码列表模块id、预测模块id和回测模块id,如下图所示:

第三步:

设置数据开始日期start_date,更新周期train_update_days,最小数据天数train_data_min_days和最大数据天数train_data_max_days,如下图所示:

这里我们以每250个交易日构建并更新一次模型来预测未来250个交易日为例,即:train_data_max_days=250,train_data_min_days=250,train_update_days=250

需要注意的是设置的自动标注shift(close, -N) / shift(open, -1)中训练集前N日的标注为Nan,因此N的取值要小于train_data_min_days,否则会因为训练集标注没有数据报错。

第四步:

运行策略,完整的滚动训练案例如下所示:

提示:滚动训练属于高级优化模块,其调用的是整个可视化画布,因此不能单独运行该模块,而是 点击运行全部

克隆策略

    {"Description":"实验创建于2017/8/26","Summary":"","Graph":{"EdgesInternal":[{"DestinationInputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-15:instruments","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-8:data"},{"DestinationInputPortId":"-560:instruments","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-8:data"},{"DestinationInputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-53:data1","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-15:data"},{"DestinationInputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-43:features","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"DestinationInputPortId":"-560:features","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"DestinationInputPortId":"-567:features","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"DestinationInputPortId":"-574:features","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"DestinationInputPortId":"-581:features","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-24:data"},{"DestinationInputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-60:model","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-43:model"},{"DestinationInputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-84:input_data","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-53:data"},{"DestinationInputPortId":"-591:options_data","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-60:predictions"},{"DestinationInputPortId":"-574:instruments","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-62:data"},{"DestinationInputPortId":"-591:instruments","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-62:data"},{"DestinationInputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-43:training_ds","SourceOutputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-84:data"},{"DestinationInputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-60:data","SourceOutputPortId":"-86:data"},{"DestinationInputPortId":"-567:input_data","SourceOutputPortId":"-560:data"},{"DestinationInputPortId":"287d2cb0-f53c-4101-bdf8-104b137c8601-53:data2","SourceOutputPortId":"-567:data"},{"DestinationInputPortId":"-581:input_data","SourceOutputPortId":"-574:data"},{"DestinationInputPortId":"-86:input_data","SourceOutputPortId":"-581:data"}],"ModuleNodes":[{"Id":"287d2cb0-f53c-4101-bdf8-104b137c8601-8","ModuleId":"BigQuantSpace.instruments.instruments-v2","ModuleParameters":[{"Name":"start_date","Value":"2010-01-01","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"end_date","Value":"2015-01-01","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"market","Value":"CN_STOCK_A","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"instrument_list","Value":"","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"max_count","Value":"0","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"rolling_conf","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-8"}],"OutputPortsInternal":[{"Name":"data","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-8","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":1,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"287d2cb0-f53c-4101-bdf8-104b137c8601-15","ModuleId":"BigQuantSpace.advanced_auto_labeler.advanced_auto_labeler-v2","ModuleParameters":[{"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, -22) / shift(open, -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# 将分数映射到分类,这里使用20个分类\nall_wbins(label, 20)\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","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"start_date","Value":"","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"end_date","Value":"","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"benchmark","Value":"000300.SHA","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"drop_na_label","Value":"True","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"cast_label_int","Value":"True","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"user_functions","Value":"","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"instruments","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-15"}],"OutputPortsInternal":[{"Name":"data","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-15","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":2,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"287d2cb0-f53c-4101-bdf8-104b137c8601-24","ModuleId":"BigQuantSpace.input_features.input_features-v1","ModuleParameters":[{"Name":"features","Value":"# #号开始的表示注释\n# 多个特征,每行一个,可以包含基础特征和衍生特征\nreturn_5\nreturn_10\nreturn_20\navg_amount_0/avg_amount_5\navg_amount_5/avg_amount_20\nrank_avg_amount_0/rank_avg_amount_5\nrank_avg_amount_5/rank_avg_amount_10\nrank_return_0\nrank_return_5\nrank_return_10\nrank_return_0/rank_return_5\nrank_return_5/rank_return_10\npe_ttm_0","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"features_ds","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-24"}],"OutputPortsInternal":[{"Name":"data","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-24","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":3,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"287d2cb0-f53c-4101-bdf8-104b137c8601-43","ModuleId":"BigQuantSpace.stock_ranker_train.stock_ranker_train-v5","ModuleParameters":[{"Name":"learning_algorithm","Value":"排序","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"number_of_leaves","Value":30,"ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"minimum_docs_per_leaf","Value":1000,"ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"number_of_trees","Value":20,"ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"learning_rate","Value":0.1,"ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"max_bins","Value":1023,"ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"feature_fraction","Value":1,"ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"m_lazy_run","Value":"False","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"training_ds","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-43"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"features","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-43"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"test_ds","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-43"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"base_model","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-43"}],"OutputPortsInternal":[{"Name":"model","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-43","OutputType":null},{"Name":"feature_gains","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-43","OutputType":null},{"Name":"m_lazy_run","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-43","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":6,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"287d2cb0-f53c-4101-bdf8-104b137c8601-53","ModuleId":"BigQuantSpace.join.join-v3","ModuleParameters":[{"Name":"on","Value":"date,instrument","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"how","Value":"inner","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"sort","Value":"False","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"data1","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-53"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"data2","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-53"}],"OutputPortsInternal":[{"Name":"data","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-53","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":7,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"287d2cb0-f53c-4101-bdf8-104b137c8601-60","ModuleId":"BigQuantSpace.stock_ranker_predict.stock_ranker_predict-v5","ModuleParameters":[{"Name":"m_lazy_run","Value":"False","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"model","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-60"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"data","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-60"}],"OutputPortsInternal":[{"Name":"predictions","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-60","OutputType":null},{"Name":"m_lazy_run","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-60","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":8,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"287d2cb0-f53c-4101-bdf8-104b137c8601-62","ModuleId":"BigQuantSpace.instruments.instruments-v2","ModuleParameters":[{"Name":"start_date","Value":"2017-01-01","ValueType":"Literal","LinkedGlobalParameter":"交易日期"},{"Name":"end_date","Value":"2018-10-22","ValueType":"Literal","LinkedGlobalParameter":"交易日期"},{"Name":"market","Value":"CN_STOCK_A","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"instrument_list","Value":"","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"max_count","Value":"0","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"rolling_conf","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-62"}],"OutputPortsInternal":[{"Name":"data","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-62","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":9,"IsPartOfPartialRun":null,"Comment":"预测数据,用于回测和模拟","CommentCollapsed":false},{"Id":"287d2cb0-f53c-4101-bdf8-104b137c8601-84","ModuleId":"BigQuantSpace.dropnan.dropnan-v1","ModuleParameters":[],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"input_data","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-84"}],"OutputPortsInternal":[{"Name":"data","NodeId":"287d2cb0-f53c-4101-bdf8-104b137c8601-84","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":13,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"-86","ModuleId":"BigQuantSpace.dropnan.dropnan-v1","ModuleParameters":[],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"input_data","NodeId":"-86"}],"OutputPortsInternal":[{"Name":"data","NodeId":"-86","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":14,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"-560","ModuleId":"BigQuantSpace.general_feature_extractor.general_feature_extractor-v7","ModuleParameters":[{"Name":"start_date","Value":"","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"end_date","Value":"","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"before_start_days","Value":"90","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"instruments","NodeId":"-560"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"features","NodeId":"-560"}],"OutputPortsInternal":[{"Name":"data","NodeId":"-560","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":15,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"-567","ModuleId":"BigQuantSpace.derived_feature_extractor.derived_feature_extractor-v3","ModuleParameters":[{"Name":"date_col","Value":"date","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"instrument_col","Value":"instrument","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"drop_na","Value":"False","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"remove_extra_columns","Value":"False","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"user_functions","Value":"","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"input_data","NodeId":"-567"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"features","NodeId":"-567"}],"OutputPortsInternal":[{"Name":"data","NodeId":"-567","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":16,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"-574","ModuleId":"BigQuantSpace.general_feature_extractor.general_feature_extractor-v7","ModuleParameters":[{"Name":"start_date","Value":"","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"end_date","Value":"","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"before_start_days","Value":"90","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"instruments","NodeId":"-574"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"features","NodeId":"-574"}],"OutputPortsInternal":[{"Name":"data","NodeId":"-574","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":17,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"-581","ModuleId":"BigQuantSpace.derived_feature_extractor.derived_feature_extractor-v3","ModuleParameters":[{"Name":"date_col","Value":"date","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"instrument_col","Value":"instrument","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"drop_na","Value":"False","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"remove_extra_columns","Value":"False","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"user_functions","Value":"","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"input_data","NodeId":"-581"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"features","NodeId":"-581"}],"OutputPortsInternal":[{"Name":"data","NodeId":"-581","OutputType":null}],"UsePreviousResults":true,"moduleIdForCode":18,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"-591","ModuleId":"BigQuantSpace.trade.trade-v4","ModuleParameters":[{"Name":"start_date","Value":"","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"end_date","Value":"","ValueType":"Literal","LinkedGlobalParameter":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 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","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"prepare","Value":"# 回测引擎:准备数据,只执行一次\ndef bigquant_run(context):\n pass\n","ValueType":"Literal","LinkedGlobalParameter":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=0.0003, sell_cost=0.0013, 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 = 30\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'] = 22\n","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"before_trading_start","Value":"","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"volume_limit","Value":0.025,"ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"order_price_field_buy","Value":"open","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"order_price_field_sell","Value":"close","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"capital_base","Value":1000000,"ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"auto_cancel_non_tradable_orders","Value":"True","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"data_frequency","Value":"daily","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"price_type","Value":"后复权","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"product_type","Value":"股票","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"plot_charts","Value":"True","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"backtest_only","Value":"False","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"benchmark","Value":"","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"instruments","NodeId":"-591"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"options_data","NodeId":"-591"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"history_ds","NodeId":"-591"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"benchmark_ds","NodeId":"-591"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"trading_calendar","NodeId":"-591"}],"OutputPortsInternal":[{"Name":"raw_perf","NodeId":"-591","OutputType":null}],"UsePreviousResults":false,"moduleIdForCode":19,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true},{"Id":"-613","ModuleId":"BigQuantSpace.hyper_rolling_train.hyper_rolling_train-v1","ModuleParameters":[{"Name":"run","Value":"def bigquant_run(\n bq_graph,\n inputs,\n trading_days_market='CN', # 使用那个市场的交易日历\n train_instruments_mid='m1', # 训练数据 证券代码列表 模块id\n test_instruments_mid='m9', # 测试数据 证券代码列表 模块id\n predict_mid='m8', # 预测 模块id\n trade_mid='m19', # 回测 模块id\n start_date='2016-01-01', # 数据开始日期\n end_date=T.live_run_param('trading_date', '2018-10-22'), # 数据结束日期\n train_update_days=250, # 更新周期,按交易日计算,每多少天更新一次\n train_update_days_for_live=None, #模拟实盘模式下的更新周期,按交易日计算,每多少天更新一次。如果需要在模拟实盘阶段使用不同的模型更新周期,可以设置这个参数\n train_data_min_days=250, # 最小数据天数,按交易日计算,所以第一个滚动的结束日期是 从开始日期到开始日期+最小数据天数\n train_data_max_days=250, # 最大数据天数,按交易日计算,0,表示没有限制,否则每一个滚动的开始日期=max(此滚动的结束日期-最大数据天数, 开始日期\n rolling_count_for_live=1, #实盘模式下滚动次数,模拟实盘模式下,取最后多少次滚动。一般在模拟实盘模式下,只用到最后一次滚动训练的模型,这里可以设置为1;如果你的滚动训练数据时间段很短,以至于期间可能没有训练数据,这里可以设置大一点。0表示没有限制\n):\n def merge_datasources(input_1):\n df_list = [ds[0].read_df().set_index('date').ix[ds[1]:].reset_index() for ds in input_1]\n df = pd.concat(df_list)\n instrument_data = {\n 'start_date': df['date'].min().strftime('%Y-%m-%d'),\n 'end_date': df['date'].max().strftime('%Y-%m-%d'),\n 'instruments': list(set(df['instrument'])),\n }\n return Outputs(data=DataSource.write_df(df), instrument_data=DataSource.write_pickle(instrument_data))\n\n def gen_rolling_dates(trading_days_market, start_date, end_date, train_update_days, train_update_days_for_live, train_data_min_days, train_data_max_days, rolling_count_for_live):\n # 是否实盘模式\n tdays = list(D.trading_days(market=trading_days_market, start_date=start_date, end_date=end_date)['date'])\n is_live_run = T.live_run_param('trading_date', None) is not None\n\n if is_live_run and train_update_days_for_live:\n train_update_days = train_update_days_for_live\n\n rollings = []\n train_end_date = train_data_min_days\n while train_end_date < len(tdays):\n if train_data_max_days is not None:\n train_start_date = max(train_end_date - train_data_max_days, 0)\n else:\n train_start_date = start_date\n rollings.append({\n 'train_start_date': tdays[train_start_date].strftime('%Y-%m-%d'),\n 'train_end_date': tdays[train_end_date - 1].strftime('%Y-%m-%d'),\n 'test_start_date': tdays[train_end_date].strftime('%Y-%m-%d'),\n 'test_end_date': tdays[min(train_end_date + train_update_days, len(tdays)) - 1].strftime('%Y-%m-%d'),\n })\n train_end_date += train_update_days\n\n if not rollings:\n raise Exception('没有滚动需要执行,请检查配置')\n\n if is_live_run and rolling_count_for_live:\n rollings = rollings[-rolling_count_for_live:]\n\n return rollings\n\n g = bq_graph\n\n rolling_dates = gen_rolling_dates(\n trading_days_market, start_date, end_date, train_update_days, train_update_days_for_live, train_data_min_days, train_data_max_days, rolling_count_for_live)\n\n # 训练和预测\n results = []\n for rolling in rolling_dates:\n parameters = {}\n # 先禁用回测\n parameters[trade_mid + '.__enabled__'] = False\n parameters[train_instruments_mid + '.start_date'] = rolling['train_start_date']\n parameters[train_instruments_mid + '.end_date'] = rolling['train_end_date']\n parameters[test_instruments_mid + '.start_date'] = rolling['test_start_date']\n parameters[test_instruments_mid + '.end_date'] = rolling['test_end_date']\n # print('------ rolling_train:', parameters)\n results.append(g.run(parameters))\n\n # 合并预测结果并回测\n mx = M.cached.v3(run=merge_datasources, input_1=[[result[predict_mid].predictions,result[test_instruments_mid].data.read_pickle()['start_date']] for result in results])\n parameters = {}\n parameters['*.__enabled__'] = False\n parameters[trade_mid + '.__enabled__'] = True\n parameters[trade_mid + '.instruments'] = mx.instrument_data\n parameters[trade_mid + '.options_data'] = mx.data\n\n trade = g.run(parameters)\n\n return {'rollings': results, 'trade': trade}\n","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"run_now","Value":"True","ValueType":"Literal","LinkedGlobalParameter":null},{"Name":"bq_graph","Value":"","ValueType":"Literal","LinkedGlobalParameter":null}],"InputPortsInternal":[{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"bq_graph_port","NodeId":"-613"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"input_1","NodeId":"-613"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"input_2","NodeId":"-613"},{"DataSourceId":null,"TrainedModelId":null,"TransformModuleId":null,"Name":"input_3","NodeId":"-613"}],"OutputPortsInternal":[{"Name":"result","NodeId":"-613","OutputType":null}],"UsePreviousResults":false,"moduleIdForCode":20,"IsPartOfPartialRun":null,"Comment":"","CommentCollapsed":true}],"SerializedClientData":"<?xml version='1.0' encoding='utf-16'?><DataV1 xmlns:xsd='http://www.w3.org/2001/XMLSchema' xmlns:xsi='http://www.w3.org/2001/XMLSchema-instance'><Meta /><NodePositions><NodePosition Node='287d2cb0-f53c-4101-bdf8-104b137c8601-8' Position='207,75,200,200'/><NodePosition Node='287d2cb0-f53c-4101-bdf8-104b137c8601-15' Position='70,194,200,200'/><NodePosition Node='287d2cb0-f53c-4101-bdf8-104b137c8601-24' Position='585,-25,200,200'/><NodePosition Node='287d2cb0-f53c-4101-bdf8-104b137c8601-43' Position='639,561,200,200'/><NodePosition Node='287d2cb0-f53c-4101-bdf8-104b137c8601-53' Position='249,375,200,200'/><NodePosition Node='287d2cb0-f53c-4101-bdf8-104b137c8601-60' Position='906,647,200,200'/><NodePosition Node='287d2cb0-f53c-4101-bdf8-104b137c8601-62' Position='1071,125,200,200'/><NodePosition Node='287d2cb0-f53c-4101-bdf8-104b137c8601-84' Position='376,467,200,200'/><NodePosition Node='-86' Position='1078,418,200,200'/><NodePosition Node='-560' Position='381,188,200,200'/><NodePosition Node='-567' Position='385,280,200,200'/><NodePosition Node='-574' Position='1078,236,200,200'/><NodePosition Node='-581' Position='1081,327,200,200'/><NodePosition Node='-591' Position='1033,749,200,200'/><NodePosition Node='-613' Position='268,739,200,200'/></NodePositions><NodeGroups /></DataV1>"},"IsDraft":true,"ParentExperimentId":null,"WebService":{"IsWebServiceExperiment":false,"Inputs":[],"Outputs":[],"Parameters":[{"Name":"交易日期","Value":"","ParameterDefinition":{"Name":"交易日期","FriendlyName":"交易日期","DefaultValue":"","ParameterType":"String","HasDefaultValue":true,"IsOptional":true,"ParameterRules":[],"HasRules":false,"MarkupType":0,"CredentialDescriptor":null}}],"WebServiceGroupId":null,"SerializedClientData":"<?xml version='1.0' encoding='utf-16'?><DataV1 xmlns:xsd='http://www.w3.org/2001/XMLSchema' xmlns:xsi='http://www.w3.org/2001/XMLSchema-instance'><Meta /><NodePositions></NodePositions><NodeGroups /></DataV1>"},"DisableNodesUpdate":false,"Category":"user","Tags":[],"IsPartialRun":true}
    In [1]:
    # 本代码由可视化策略环境自动生成 2019年3月20日 09:44
    # 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。
    
    
    # 回测引擎:每日数据处理函数,每天执行一次
    def m19_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]))])))
            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 m19_prepare_bigquant_run(context):
        pass
    
    # 回测引擎:初始化函数,只执行一次
    def m19_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=0.0003, sell_cost=0.0013, 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 = 30
        # 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'] = 22
    
    
    g = T.Graph({
    
        'm1': 'M.instruments.v2',
        'm1.start_date': '2010-01-01',
        'm1.end_date': '2015-01-01',
        'm1.market': 'CN_STOCK_A',
        'm1.instrument_list': '',
        'm1.max_count': 0,
    
        'm2': 'M.advanced_auto_labeler.v2',
        'm2.instruments': T.Graph.OutputPort('m1.data'),
        'm2.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, -22) / shift(open, -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))
    
    # 将分数映射到分类,这里使用20个分类
    all_wbins(label, 20)
    
    # 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)
    """,
        'm2.start_date': '',
        'm2.end_date': '',
        'm2.benchmark': '000300.SHA',
        'm2.drop_na_label': True,
        'm2.cast_label_int': True,
    
        'm3': 'M.input_features.v1',
        'm3.features': """# #号开始的表示注释
    # 多个特征,每行一个,可以包含基础特征和衍生特征
    return_5
    return_10
    return_20
    avg_amount_0/avg_amount_5
    avg_amount_5/avg_amount_20
    rank_avg_amount_0/rank_avg_amount_5
    rank_avg_amount_5/rank_avg_amount_10
    rank_return_0
    rank_return_5
    rank_return_10
    rank_return_0/rank_return_5
    rank_return_5/rank_return_10
    pe_ttm_0""",
    
        'm15': 'M.general_feature_extractor.v7',
        'm15.instruments': T.Graph.OutputPort('m1.data'),
        'm15.features': T.Graph.OutputPort('m3.data'),
        'm15.start_date': '',
        'm15.end_date': '',
        'm15.before_start_days': 90,
    
        'm16': 'M.derived_feature_extractor.v3',
        'm16.input_data': T.Graph.OutputPort('m15.data'),
        'm16.features': T.Graph.OutputPort('m3.data'),
        'm16.date_col': 'date',
        'm16.instrument_col': 'instrument',
        'm16.drop_na': False,
        'm16.remove_extra_columns': False,
    
        'm7': 'M.join.v3',
        'm7.data1': T.Graph.OutputPort('m2.data'),
        'm7.data2': T.Graph.OutputPort('m16.data'),
        'm7.on': 'date,instrument',
        'm7.how': 'inner',
        'm7.sort': False,
    
        'm13': 'M.dropnan.v1',
        'm13.input_data': T.Graph.OutputPort('m7.data'),
    
        'm6': 'M.stock_ranker_train.v5',
        'm6.training_ds': T.Graph.OutputPort('m13.data'),
        'm6.features': T.Graph.OutputPort('m3.data'),
        'm6.learning_algorithm': '排序',
        'm6.number_of_leaves': 30,
        'm6.minimum_docs_per_leaf': 1000,
        'm6.number_of_trees': 20,
        'm6.learning_rate': 0.1,
        'm6.max_bins': 1023,
        'm6.feature_fraction': 1,
        'm6.m_lazy_run': False,
    
        'm9': 'M.instruments.v2',
        'm9.start_date': T.live_run_param('trading_date', '2017-01-01'),
        'm9.end_date': T.live_run_param('trading_date', '2018-10-22'),
        'm9.market': 'CN_STOCK_A',
        'm9.instrument_list': '',
        'm9.max_count': 0,
    
        'm17': 'M.general_feature_extractor.v7',
        'm17.instruments': T.Graph.OutputPort('m9.data'),
        'm17.features': T.Graph.OutputPort('m3.data'),
        'm17.start_date': '',
        'm17.end_date': '',
        'm17.before_start_days': 90,
    
        'm18': 'M.derived_feature_extractor.v3',
        'm18.input_data': T.Graph.OutputPort('m17.data'),
        'm18.features': T.Graph.OutputPort('m3.data'),
        'm18.date_col': 'date',
        'm18.instrument_col': 'instrument',
        'm18.drop_na': False,
        'm18.remove_extra_columns': False,
    
        'm14': 'M.dropnan.v1',
        'm14.input_data': T.Graph.OutputPort('m18.data'),
    
        'm8': 'M.stock_ranker_predict.v5',
        'm8.model': T.Graph.OutputPort('m6.model'),
        'm8.data': T.Graph.OutputPort('m14.data'),
        'm8.m_lazy_run': False,
    
        'm19': 'M.trade.v4',
        'm19.instruments': T.Graph.OutputPort('m9.data'),
        'm19.options_data': T.Graph.OutputPort('m8.predictions'),
        'm19.start_date': '',
        'm19.end_date': '',
        'm19.handle_data': m19_handle_data_bigquant_run,
        'm19.prepare': m19_prepare_bigquant_run,
        'm19.initialize': m19_initialize_bigquant_run,
        'm19.volume_limit': 0.025,
        'm19.order_price_field_buy': 'open',
        'm19.order_price_field_sell': 'close',
        'm19.capital_base': 1000000,
        'm19.auto_cancel_non_tradable_orders': True,
        'm19.data_frequency': 'daily',
        'm19.price_type': '后复权',
        'm19.product_type': '股票',
        'm19.plot_charts': True,
        'm19.backtest_only': False,
        'm19.benchmark': '',
    })
    
    # g.run({})
    
    
    def m20_run_bigquant_run(
        bq_graph,
        inputs,
        trading_days_market='CN', # 使用那个市场的交易日历
        train_instruments_mid='m1', # 训练数据 证券代码列表 模块id
        test_instruments_mid='m9', # 测试数据 证券代码列表 模块id
        predict_mid='m8', # 预测 模块id
        trade_mid='m19', # 回测 模块id
        start_date='2016-01-01', # 数据开始日期
        end_date=T.live_run_param('trading_date', '2018-10-22'), # 数据结束日期
        train_update_days=250, # 更新周期,按交易日计算,每多少天更新一次
        train_update_days_for_live=None, #模拟实盘模式下的更新周期,按交易日计算,每多少天更新一次。如果需要在模拟实盘阶段使用不同的模型更新周期,可以设置这个参数
        train_data_min_days=250, # 最小数据天数,按交易日计算,所以第一个滚动的结束日期是 从开始日期到开始日期+最小数据天数
        train_data_max_days=250, # 最大数据天数,按交易日计算,0,表示没有限制,否则每一个滚动的开始日期=max(此滚动的结束日期-最大数据天数, 开始日期
        rolling_count_for_live=1, #实盘模式下滚动次数,模拟实盘模式下,取最后多少次滚动。一般在模拟实盘模式下,只用到最后一次滚动训练的模型,这里可以设置为1;如果你的滚动训练数据时间段很短,以至于期间可能没有训练数据,这里可以设置大一点。0表示没有限制
    ):
        def merge_datasources(input_1):
            df_list = [ds[0].read_df().set_index('date').ix[ds[1]:].reset_index() for ds in input_1]
            df = pd.concat(df_list)
            instrument_data = {
                'start_date': df['date'].min().strftime('%Y-%m-%d'),
                'end_date': df['date'].max().strftime('%Y-%m-%d'),
                'instruments': list(set(df['instrument'])),
            }
            return Outputs(data=DataSource.write_df(df), instrument_data=DataSource.write_pickle(instrument_data))
    
        def gen_rolling_dates(trading_days_market, start_date, end_date, train_update_days, train_update_days_for_live, train_data_min_days, train_data_max_days, rolling_count_for_live):
            # 是否实盘模式
            tdays = list(D.trading_days(market=trading_days_market, start_date=start_date, end_date=end_date)['date'])
            is_live_run = T.live_run_param('trading_date', None) is not None
    
            if is_live_run and train_update_days_for_live:
                train_update_days = train_update_days_for_live
    
            rollings = []
            train_end_date = train_data_min_days
            while train_end_date < len(tdays):
                if train_data_max_days is not None:
                    train_start_date = max(train_end_date - train_data_max_days, 0)
                else:
                    train_start_date = start_date
                rollings.append({
                    'train_start_date': tdays[train_start_date].strftime('%Y-%m-%d'),
                    'train_end_date': tdays[train_end_date - 1].strftime('%Y-%m-%d'),
                    'test_start_date': tdays[train_end_date].strftime('%Y-%m-%d'),
                    'test_end_date': tdays[min(train_end_date + train_update_days, len(tdays)) - 1].strftime('%Y-%m-%d'),
                })
                train_end_date += train_update_days
    
            if not rollings:
                raise Exception('没有滚动需要执行,请检查配置')
    
            if is_live_run and rolling_count_for_live:
                rollings = rollings[-rolling_count_for_live:]
    
            return rollings
    
        g = bq_graph
    
        rolling_dates = gen_rolling_dates(
            trading_days_market, start_date, end_date, train_update_days, train_update_days_for_live, train_data_min_days, train_data_max_days, rolling_count_for_live)
    
        # 训练和预测
        results = []
        for rolling in rolling_dates:
            parameters = {}
            # 先禁用回测
            parameters[trade_mid + '.__enabled__'] = False
            parameters[train_instruments_mid + '.start_date'] = rolling['train_start_date']
            parameters[train_instruments_mid + '.end_date'] = rolling['train_end_date']
            parameters[test_instruments_mid + '.start_date'] = rolling['test_start_date']
            parameters[test_instruments_mid + '.end_date'] = rolling['test_end_date']
            # print('------ rolling_train:', parameters)
            results.append(g.run(parameters))
    
        # 合并预测结果并回测
        mx = M.cached.v3(run=merge_datasources, input_1=[[result[predict_mid].predictions,result[test_instruments_mid].data.read_pickle()['start_date']] for result in results])
        parameters = {}
        parameters['*.__enabled__'] = False
        parameters[trade_mid + '.__enabled__'] = True
        parameters[trade_mid + '.instruments'] = mx.instrument_data
        parameters[trade_mid + '.options_data'] = mx.data
    
        trade = g.run(parameters)
    
        return {'rollings': results, 'trade': trade}
    
    
    m20 = M.hyper_rolling_train.v1(
        run=m20_run_bigquant_run,
        run_now=True,
        bq_graph=g
    )
    
    • 收益率-44.67%
    • 年化收益率-29.2%
    • 基准收益率-2.62%
    • 阿尔法-0.33
    • 贝塔0.85
    • 夏普比率-1.91
    • 胜率0.47
    • 盈亏比0.53
    • 收益波动率18.7%
    • 信息比率-0.16
    • 最大回撤50.49%

    我们通过滚动训练模块的结果查看各次滚动训练状态

    # 第一次训练的训练集
    print(m20.result['rollings'][0]['m1'].data.read_pickle()['start_date'])
    print(m20.result['rollings'][0]['m1'].data.read_pickle()['end_date'])
    # 第一次训练的模型的特征得分
    m20.result['rollings'][0]['m6'].feature_gains.read_df()
    # 第一次训练的预测集
    print(m20.result['rollings'][0]['m9'].data.read_pickle()['start_date'])
    print(m20.result['rollings'][0]['m9'].data.read_pickle()['end_date'])
    # 第一次训练的预测集
    print(m20.result['rollings'][0]['m9'].data.read_pickle()['start_date'])
    print(m20.result['rollings'][0]['m9'].data.read_pickle()['end_date'])
    # 训练的第一个模型的预测结果
    m20.result['rollings'][0]['m8'].predictions.read_df().set_index('date').ix['2017-02-14'].head()
    # 第二次训练的训练集
    print(m20.result['rollings'][1]['m1'].data.read_pickle()['start_date'])
    print(m20.result['rollings'][1]['m1'].data.read_pickle()['end_date'])
    

    三、滚动训练流程解读

    在滚动训练中,我们设置了起始日期和模型更新间隔以及train_data_min_days和train_data_max_days后:
    第一步,由gen_rolling_dates函数生成一系列滚动的时间段,每个时间段的起止日期以字典格式记录,例如{‘train_start_date’:‘2010-01-04’,‘train_end_date’:‘2011-01-06’]}。不同的时间段字典记录在rolling_dates列表中。

    第二步,在每个时间段中循环运行策略,每个时间段的运行过程中禁用跳过trade模块,运行到预测模块就结束,并将每个时间段的运行结果记录在results列表中。

    实现代码
        # 训练和预测
        results = []
        for rolling in rolling_dates:
            parameters = {}
            # 先禁用回测
            parameters[trade_mid + '.__enabled__'] = False
            parameters[train_instruments_mid + '.start_date'] = rolling['train_start_date']
            parameters[train_instruments_mid + '.end_date'] = rolling['train_end_date']
            parameters[test_instruments_mid + '.start_date'] = rolling['test_start_date']
            parameters[test_instruments_mid + '.end_date'] = rolling['test_end_date']
            # print('------ rolling_train:', parameters)
            results.append(g.run(parameters))
    

    第三步,我们通过merge_datasources函数将每个时间段运行结果中的预测模块结果做拼接,输出的每日预测结果和回测总的起止时间和股票列表。

    第四步,将合并后的结果传递给Trade回测模块获得最终的滚动回测结果。

    通过滚动训练中的train_data_min_days、train_data_max_days和train_update_days参数可以轻松实现模型的训练集的滚动更新,让你的模型根据行情变化动起来~


    如何使用滚动训练并查看因子影响力(feature gains)
    滚动训练报错‘DataSourse’ object dose not support indexing,怎么办?
    (tkyz) #2

    新手,滚动训练可视图怎么连?


    (iQuant) #3

    不需要连,拖进来就可以了。可以根据需要配置参数,进一步可以修改滚动训练的代码。

    image


    (冰柠檬) #4

    [ start_date=‘2016-01-01’, # 数据开始日期
    end_date=T.live_run_param(‘trading_date’, ‘2018-10-22’), # 数据结束日期

    这两个日期指的是回测日期吗?如果不是 是不是应该把测试集都包含进去比较好呢?


    (iQuant) #5

    这是整个实验段的起始日期。模块会根据参数生成有训练和测试数据的rollings。具体请参考代码。可以在代码里把 rolling_dates 输出出来看看。


    (tkyz) #6

    train_update_days_for_live,赋值为None的时候,是实盘的更新周期与回测的更新周期一样吗?


    (iQuant) #7

    您好,是一样的。


    (focus666) #8

    加入滚动训练后 训练集和测试集的日期是不是就失效了? 假如 我的训练集数据是2010-01-01到2018-01-01 我的测试测试集数据是2018-01-02到2019-01-29 我想每90天更新一次 这个开始日期和结束日期应该怎么设置?


    (达达) #9

    加了滚动模块后,时间段由滚动训练模块确定,原有的证券代码列表中的时间无效了


    (xuan) #10

    不用连线方便了不少,只是的好好看看了。


    (yangziriver) #11

    克隆策略运行后出现错误


    (iQuant) #12

    您是有什么修改吗?这边跑是没问题的,重启内核试试呢。


    (yangziriver) #13

    我退出了浏览器重新进入,重新克隆策略,运行后又有新的错了,是m1,这之前还有一次是m19.


    (iQuant) #14

    有可能是缓存的问题,您使用的是什么浏览器呢?建议您使用chrome浏览器


    (yangziriver) #15

    是用的google浏览器。


    (yangziriver) #16

    原因是从 本贴克隆的策略,滚动训练模板出现红点image
    从训练营老师处下载的策略中的滚动训练模板没有问题
    Uploading: image.png…
    有问题的标注m20,没问题的标注m4。代码我对照过了,没有其它差别。只好直接把m4复制过去了。
    image


    (yangziriver) #17

    image
    7、滚动训练模块是好的,从陈老师的课件复制过来的,滚动训练功能还是不能滚动训练,和原来一样的问题,模块上红点没有了。我仔细对照了python代码,没有差别。真是怪事。是不是服务器认策略名,造成了冲突?


    (iQuant) #18

    我们再看一下。


    (cash01) #20

    还有结果吗?我的也报错


    (iQuant) #21

    稍等 帮您确定一下。