复制链接

克隆策略

In [40]:

# 本代码由可视化策略环境自动生成 2023年2月16日 16:54
# 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。


def talib_KDJ(df,high, low, close, fastk_period=9, slowk_period=3, slowd_period=3):
    #计算kd指标
    k, d = talib.STOCH(high, low, close, 
                                       fastk_period=fastk_period, 
                                       slowk_period=slowk_period, 
                                       slowd_period=slowd_period)
    j = 3 * k - 2 *d
    return j 

# 因为这俩表达式是我们新自定义的表达式，因此需要声明，以便在输入特征列表可使用
m18_user_functions_bigquant_run = {
    'talib_KDJ_j':talib_KDJ
}
def talib_KDJ(df,high, low, close, fastk_period=9, slowk_period=3, slowd_period=3):
    #计算kd指标
    k, d = talib.STOCH(high, low, close, 
                                       fastk_period=fastk_period, 
                                       slowk_period=slowk_period, 
                                       slowd_period=slowd_period)
    j = 3 * k - 2 *d
    return j 

# 因为这俩表达式是我们新自定义的表达式，因此需要声明，以便在输入特征列表可使用
m25_user_functions_bigquant_run = {
    'talib_KDJ_j':talib_KDJ
}
def talib_KDJ(df,high, low, close, fastk_period=9, slowk_period=3, slowd_period=3):
    #计算kd指标
    k, d = talib.STOCH(high, low, close, 
                                       fastk_period=fastk_period, 
                                       slowk_period=slowk_period, 
                                       slowd_period=slowd_period)
    j = 3 * k - 2 *d
    return j 

# 因为这俩表达式是我们新自定义的表达式，因此需要声明，以便在输入特征列表可使用
m16_user_functions_bigquant_run = {
    'talib_KDJ_j':talib_KDJ
}
def talib_KDJ(df,high, low, close, fastk_period=9, slowk_period=3, slowd_period=3):
    #计算kd指标
    k, d = talib.STOCH(high, low, close, 
                                       fastk_period=fastk_period, 
                                       slowk_period=slowk_period, 
                                       slowd_period=slowd_period)
    j = 3 * k - 2 *d
    return j 

# 因为这俩表达式是我们新自定义的表达式，因此需要声明，以便在输入特征列表可使用
m32_user_functions_bigquant_run = {
    'talib_KDJ_j':talib_KDJ
}
# 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 = 5
    # 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 = 0.34
    context.options['hold_days'] = 5

# 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):
    
    today = data.current_dt.strftime('%Y-%m-%d')
    
    #------------------------------------------止损模块START--------------------------------------------
    equities = {e.symbol: p for e, p in context.portfolio.positions.items() if p.amount>0}
    positions_stop={e.symbol:p.cost_basis for e,p in context.portfolio.positions.items()}
    
    # 新建当日止损股票列表是为了handle_data 策略逻辑部分不再对该股票进行判断
    stoploss_stock = [] 
    stopwin_stock=[]
    if len(equities) > 0:
        for i in equities.keys():
            stock_cost=positions_stop[i]
            stock_market_price = data.current(context.symbol(i), 'price')  # 最新市场价格
            last_sale_date = equities[i].last_sale_date   # 上次交易日期
            delta_days = data.current_dt - last_sale_date  
            hold_days = delta_days.days # 持仓天数
            # 建仓以来的最高价
            highest_price_since_buy = data.history(context.symbol(i), 'high', hold_days, '1d').max()
            # 确定止损位置
            stoploss_line = stock_cost * 0.9
            #record('止损位置', stoploss_line)
            # 如果价格下穿止损位置
            if stock_market_price < stoploss_line:
                context.order_target_percent(context.symbol(i), 0)     
                stoploss_stock.append(i)
            if stock_market_price/stock_cost > 1.06 and data.can_trade(context.symbol(i)) and not context.has_unfinished_sell_order(i):
                context.order_target_percent(context.symbol(i),0)      
                stopwin_stock.append(i)
        if len(stoploss_stock)>0:
            print('日期：', today, '股票：', stoploss_stock, '出现跟踪止损状况')
        if len(stopwin_stock)>0:
            print(today,'止盈股票列表',stopwin_stock)
    #-------------------------------------------止损模块END---------------------------------------------    
    
    # 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()}

    sell_stock = []
#     # 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:
#             # 如果已经移动止损卖出过则不再轮仓卖出,以防止出现空头持仓
#             if instrument in stoploss_stock:
#                 continue
#             context.order_target(context.symbol(instrument), 0)
#             cash_for_sell -= positions[instrument]
#             # 记录轮仓卖出的股票
#             sell_stock.append(instrument)
#             if cash_for_sell <= 0:
#                 break
#             print(today,'末位淘汰',sell_stock)
    # 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_list = list(ranker_prediction.instrument)
    # 不再买入已经轮仓卖出和移动止损的股票,以防止出现空头持仓
    buy_instruments = [i for i in buy_list if i not in sell_stock + stoploss_stock][: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


g = T.Graph({

    'm3': 'M.input_features.v1',
    'm3.features': """return_0/return_4
pe_ttm_0

""",

    'm21': 'M.input_features.v1',
    'm21.features_ds': T.Graph.OutputPort('m3.data'),
    'm21.features': """# 未来5天的收益率
return_5_day = (shift(close_0, -5)-shift(open_0, -1))/shift(open_0, -1)

""",

    'm9': 'M.instruments.v2',
    'm9.start_date': T.live_run_param('trading_date', '2017-01-02'),
    'm9.end_date': T.live_run_param('trading_date', '2018-02-15'),
    '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('m21.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('m21.data'),
    'm18.date_col': 'date',
    'm18.instrument_col': 'instrument',
    'm18.drop_na': False,
    'm18.remove_extra_columns': False,
    'm18.user_functions': m18_user_functions_bigquant_run,

    'm27': 'M.input_features.v1',
    'm27.features': 'talib_KDJ_j(high_0,low_0,close_0)',

    'm29': 'M.instruments.v2',
    'm29.start_date': '2019-01-01',
    'm29.end_date': '2021-12-27',
    'm29.market': 'CN_STOCK_A',
    'm29.instrument_list': '',
    'm29.max_count': 0,

    'm26': 'M.general_feature_extractor.v7',
    'm26.instruments': T.Graph.OutputPort('m29.data'),
    'm26.features': T.Graph.OutputPort('m27.data'),
    'm26.start_date': '',
    'm26.end_date': '',
    'm26.before_start_days': 0,

    'm25': 'M.derived_feature_extractor.v3',
    'm25.input_data': T.Graph.OutputPort('m26.data'),
    'm25.features': T.Graph.OutputPort('m27.data'),
    'm25.date_col': 'date',
    'm25.instrument_col': 'instrument',
    'm25.drop_na': False,
    'm25.remove_extra_columns': False,
    'm25.user_functions': m25_user_functions_bigquant_run,

    'm2': 'M.advanced_auto_labeler.v2',
    'm2.instruments': T.Graph.OutputPort('m29.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/develop/datasource/deprecated/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/develop/bigexpr/usage.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))/shift(open, -1)
# product((1+shift(close, -1)/shift(open, -1)),5)
# (1+shift(close, -1)/shift(open, -1))*(1+shift(close, -2)/shift(open, -2))*(1+shift(close, -3)/shift(open, -3))*(1+shift(close, -4)/shift(open, -4))*(1+shift(close, -5)/shift(open, -5))
# where(shift(close, -2)/shift(open,-1)>1.025, 1, -1)
# where(shift(close, -3)/shift(open,-1)>1.043, 2, 0)
# where(shift(close, -4)/shift(open,-1)>1.1, 3, 0)
# rank(all_quantile(shift(close, -3)/shift(open,-1), 0.85))
rank(shift(close, -3)/shift(open,-1))
shift(close, -5) / shift(open, -1) - shift(benchmark_close, -5) / shift(benchmark_open, -1)


# 极值处理：用2%和98%分位的值做clip
clip(label, all_quantile(label, 0.02), all_quantile(label, 0.98))

# 将分数映射到分类，这里使用20个分类
all_wbins(label, 30)

# 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,

    'm15': 'M.general_feature_extractor.v7',
    'm15.instruments': T.Graph.OutputPort('m29.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,
    'm16.user_functions': m16_user_functions_bigquant_run,

    'm1': 'M.join.v3',
    'm1.data1': T.Graph.OutputPort('m25.data'),
    'm1.data2': T.Graph.OutputPort('m16.data'),
    'm1.on': 'date,instrument',
    'm1.how': 'inner',
    'm1.sort': False,

    'm7': 'M.join.v3',
    'm7.data1': T.Graph.OutputPort('m2.data'),
    'm7.data2': T.Graph.OutputPort('m1.data'),
    'm7.on': 'date,instrument',
    'm7.how': 'inner',
    'm7.sort': False,

    'm23': 'M.filtet_st_stock.v7',
    'm23.input_1': T.Graph.OutputPort('m7.data'),

    'm22': 'M.filter_stockmarket.v2',
    'm22.input_1': T.Graph.OutputPort('m23.data_1'),
    'm22.start': '3',

    'm34': 'M.filter_stockmarket.v2',
    'm34.input_1': T.Graph.OutputPort('m22.data_1'),
    'm34.start': '8',

    'm24': 'M.aa.v5',
    'm24.input_data': T.Graph.OutputPort('m34.data_1'),
    'm24.day_number': 100,

    'm13': 'M.dropnan.v1',
    'm13.input_data': T.Graph.OutputPort('m24.data'),

    'm36': 'M.stock_ranker_train.v6',
    'm36.training_ds': T.Graph.OutputPort('m13.data'),
    'm36.features': T.Graph.OutputPort('m3.data'),
    'm36.learning_algorithm': '排序',
    'm36.number_of_leaves': 30,
    'm36.minimum_docs_per_leaf': 1000,
    'm36.number_of_trees': 2,
    'm36.learning_rate': 0.1,
    'm36.max_bins': 1023,
    'm36.feature_fraction': 1,
    'm36.data_row_fraction': 1,
    'm36.plot_charts': True,
    'm36.ndcg_discount_base': 1,
    'm36.m_lazy_run': False,

    'm30': 'M.input_features.v1',
    'm30.features': 'talib_KDJ_j(high_0,low_0,close_0)',

    'm31': 'M.general_feature_extractor.v7',
    'm31.instruments': T.Graph.OutputPort('m9.data'),
    'm31.features': T.Graph.OutputPort('m30.data'),
    'm31.start_date': '',
    'm31.end_date': '',
    'm31.before_start_days': 0,

    'm32': 'M.derived_feature_extractor.v3',
    'm32.input_data': T.Graph.OutputPort('m31.data'),
    'm32.features': T.Graph.OutputPort('m30.data'),
    'm32.date_col': 'date',
    'm32.instrument_col': 'instrument',
    'm32.drop_na': False,
    'm32.remove_extra_columns': False,
    'm32.user_functions': m32_user_functions_bigquant_run,

    'm33': 'M.join.v3',
    'm33.data1': T.Graph.OutputPort('m18.data'),
    'm33.data2': T.Graph.OutputPort('m32.data'),
    'm33.on': 'date,instrument',
    'm33.how': 'inner',
    'm33.sort': False,

    'm4': 'M.filtet_st_stock.v7',
    'm4.input_1': T.Graph.OutputPort('m33.data'),

    'm5': 'M.filter_stockmarket.v2',
    'm5.input_1': T.Graph.OutputPort('m4.data_1'),
    'm5.start': '3',

    'm35': 'M.filter_stockmarket.v2',
    'm35.input_1': T.Graph.OutputPort('m5.data_1'),
    'm35.start': '8',

    'm10': 'M.aa.v5',
    'm10.input_data': T.Graph.OutputPort('m35.data_1'),
    'm10.day_number': 100,

    'm14': 'M.dropnan.v1',
    'm14.input_data': T.Graph.OutputPort('m10.data'),

    'm37': 'M.stock_ranker_predict.v5',
    'm37.model': T.Graph.OutputPort('m36.model'),
    'm37.data': T.Graph.OutputPort('m14.data'),
    'm37.m_lazy_run': False,

    'm11': 'M.join.v3',
    'm11.data1': T.Graph.OutputPort('m37.predictions'),
    'm11.data2': T.Graph.OutputPort('m14.data'),
    'm11.on': 'date,instrument',
    'm11.how': 'inner',
    'm11.sort': False,

    'm6': 'M.score_group_filter.v3',
    'm6.input_1': T.Graph.OutputPort('m11.data'),
    'm6.分组': 10,
    'm6.区间1': 10,
    'm6.区间2': 9,
    'm6.区间3': 8,

    'm8': 'M.sort.v5',
    'm8.input_ds': T.Graph.OutputPort('m6.data_1'),
    'm8.sort_by': 'score',
    'm8.group_by': 'date',
    'm8.keep_columns': '',
    'm8.ascending': True,

    'm19': 'M.trade.v4',
    'm19.instruments': T.Graph.OutputPort('m9.data'),
    'm19.options_data': T.Graph.OutputPort('m8.sorted_data'),
    'm19.start_date': '',
    'm19.end_date': '',
    'm19.initialize': m19_initialize_bigquant_run,
    'm19.handle_data': m19_handle_data_bigquant_run,
    'm19.prepare': m19_prepare_bigquant_run,
    'm19.volume_limit': 0.025,
    'm19.order_price_field_buy': 'open',
    'm19.order_price_field_sell': 'close',
    'm19.capital_base': 50000,
    'm19.auto_cancel_non_tradable_orders': True,
    'm19.data_frequency': 'daily',
    'm19.price_type': '真实价格',
    'm19.product_type': '股票',
    'm19.plot_charts': True,
    'm19.backtest_only': False,
    'm19.benchmark': '000300.SHA',
})

# g.run({})


def m28_run_bigquant_run(
    bq_graph,
    inputs,
    trading_days_market='CN', # 使用那个市场的交易日历, TODO
    train_instruments_mid='m29', # 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='m37', # 预测 模块id
    trade_mid='m19', # 回测 模块id
    model_train_mid='m6', # 模型训练 模块id
    start_date='2017-01-01', # 数据开始日期
    end_date=T.live_run_param('trading_date', '2023-02-12'), # 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=150, # 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):
        # 合并多个滚动回测的预测数据
        df_list = [ds[0].read_df().set_index('date').loc[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 and train_data_max_days > 0:
                train_start_date = max(train_end_date - train_data_max_days, 0)
            else:
                train_start_date = 0
            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)
    print('====='*5, '全部滚动的时间区间为:', rolling_dates)

    # 训练和预测
    results = []
    for rolling in rolling_dates: # 回测区间
        for nt in [10,20,30]: # number_of_trees
            for lr in [0.1,0.2]: # learning_of_rate
                parameters = {}
                # parameters[trade_mid + '.__enabled__'] = False # 将回测模块先注释
                parameters[model_train_mid + '.number_of_trees'] = nt
                parameters[model_train_mid + '.learning_rate'] = lr
                
                parameters[model_train_mid + '.plot_charts'] = False # 不绘制特征重要性图
                parameters[trade_mid + '.plot_charts'] = False # 不绘制单次滚动回测的图
                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('本次滚动回测的参数为:', nt, lr, rolling, parameters)
                
                output = {'参数详情':(nt,lr), 'graph_data':g.run(parameters)} # 将参数和图的结果保存下来
                results.append(output)
                
    # 整理预测数据
    performance_df = pd.DataFrame()

    for r in results:
        para_info = r['参数详情']
        graph_data = r['graph_data']
        test_end = graph_data[test_instruments_mid].data.read()['end_date']
        performance = graph_data[trade_mid].raw_perf.read()['sharpe'].tail(1).values[0] # 选择夏普比最大的参数组合

        tmp = pd.DataFrame({'test_end':[test_end], 'para_info':[para_info], 'performance':[performance]}) # 单次结果
        performance_df = performance_df.append(tmp)
    performance_df.index = range(len(performance_df))
    fine_paras_df = performance_df.groupby('test_end').apply(lambda x:x.sort_values('performance', ascending=False).head(1))    

    # 下列两行代码的目的：上一次滚动的最佳参数应用在下一次的预测集，因此需要关联起来，以此去找到图结果（graph_data）
    para_lst = fine_paras_df.para_info.tolist()[ :fine_paras_df.shape[0]-1] 
    test_end_lst = fine_paras_df.test_end.tolist()[-1*(fine_paras_df.shape[0]-1): ] 


    fine_para_ix = []
    for s in range(len(para_lst)):
        test_dt = test_end_lst[s]
        para = para_lst[s]
        select_df = performance_df[(performance_df['test_end'] == test_dt) & (performance_df['para_info'] == para)]
        ix = select_df.index.values[0]
        fine_para_ix.append(ix)
        print('----', ix )
    all_results = np.array(results) # 转换成array 便于批量抽取某些位置的元素
    fine_results = [r['graph_data'] for r in all_results[fine_para_ix]]
     
    # 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 fine_results])
    parameters = {}
    parameters['*.__enabled__'] = False # 将所有模块的状态设置为不运行
    parameters[trade_mid + '.__enabled__'] = True # 启动回测模块，只运行该模块
    parameters[trade_mid + '.instruments'] = mx.instrument_data
    parameters[trade_mid + '.options_data'] = mx.data
    parameters[trade_mid + '.plot_charts'] = True # 需要绘制全部结果图
    print('====='*5, '合并预测结果并回测！')

    trade = g.run(parameters)
    return {'rollings': results, 'fine_results': fine_results, 'trade': trade}


m28 = M.hyper_rolling_train.v1(
    run=m28_run_bigquant_run,
    run_now=True,
    bq_graph=g
)

[2023-02-16 16:51:35.006376] INFO: moduleinvoker: input_features.v1 开始运行..

[2023-02-16 16:51:35.063152] INFO: moduleinvoker: input_features.v1 运行完成[0.056777s].

[2023-02-16 16:51:35.067573] INFO: moduleinvoker: instruments.v2 开始运行..

[2023-02-16 16:51:35.073901] INFO: moduleinvoker: 命中缓存

[2023-02-16 16:51:35.074985] INFO: moduleinvoker: instruments.v2 运行完成[0.007414s].

[2023-02-16 16:51:35.078251] INFO: moduleinvoker: input_features.v1 开始运行..

[2023-02-16 16:51:35.087164] INFO: moduleinvoker: 命中缓存

[2023-02-16 16:51:35.088297] INFO: moduleinvoker: input_features.v1 运行完成[0.010047s].

[2023-02-16 16:51:35.092936] INFO: moduleinvoker: instruments.v2 开始运行..

[2023-02-16 16:51:35.098727] INFO: moduleinvoker: 命中缓存

[2023-02-16 16:51:35.099829] INFO: moduleinvoker: instruments.v2 运行完成[0.006893s].

[2023-02-16 16:51:35.103048] INFO: moduleinvoker: input_features.v1 开始运行..

[2023-02-16 16:51:35.107411] INFO: moduleinvoker: 命中缓存

[2023-02-16 16:51:35.108681] INFO: moduleinvoker: input_features.v1 运行完成[0.005635s].

[2023-02-16 16:51:35.112802] INFO: moduleinvoker: input_features.v1 开始运行..

[2023-02-16 16:51:35.143130] INFO: moduleinvoker: input_features.v1 运行完成[0.030298s].

[2023-02-16 16:51:35.160890] INFO: moduleinvoker: general_feature_extractor.v7 开始运行..

[2023-02-16 16:51:35.484836] INFO: 基础特征抽取: 年份 2016, 特征行数=165781

[2023-02-16 16:51:36.032518] INFO: 基础特征抽取: 年份 2017, 特征行数=443448

[2023-02-16 16:51:36.085525] INFO: 基础特征抽取: 总行数: 609229

[2023-02-16 16:51:36.092010] INFO: moduleinvoker: general_feature_extractor.v7 运行完成[0.931137s].

[2023-02-16 16:51:36.106057] INFO: moduleinvoker: general_feature_extractor.v7 开始运行..

[2023-02-16 16:51:36.115595] INFO: moduleinvoker: 命中缓存

[2023-02-16 16:51:36.117420] INFO: moduleinvoker: general_feature_extractor.v7 运行完成[0.011355s].

[2023-02-16 16:51:36.126369] INFO: moduleinvoker: advanced_auto_labeler.v2 开始运行..

[2023-02-16 16:51:36.133395] INFO: moduleinvoker: 命中缓存

[2023-02-16 16:51:36.134697] INFO: moduleinvoker: advanced_auto_labeler.v2 运行完成[0.008332s].

[2023-02-16 16:51:36.146945] INFO: moduleinvoker: general_feature_extractor.v7 开始运行..

[2023-02-16 16:51:36.154985] INFO: moduleinvoker: 命中缓存

[2023-02-16 16:51:36.156616] INFO: moduleinvoker: general_feature_extractor.v7 运行完成[0.009689s].

[2023-02-16 16:51:36.169591] INFO: moduleinvoker: general_feature_extractor.v7 开始运行..

[2023-02-16 16:51:36.686393] INFO: 基础特征抽取: 年份 2017, 特征行数=486868

[2023-02-16 16:51:37.450432] INFO: 基础特征抽取: 年份 2018, 特征行数=513321

[2023-02-16 16:51:37.566027] INFO: 基础特征抽取: 总行数: 1000189

[2023-02-16 16:51:37.571643] INFO: moduleinvoker: general_feature_extractor.v7 运行完成[1.402065s].

[2023-02-16 16:51:37.580608] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2023-02-16 16:51:38.284855] INFO: derived_feature_extractor: 提取完成 return_0/return_4, 0.004s

[2023-02-16 16:51:38.582364] INFO: derived_feature_extractor: /y_2016, 165781

[2023-02-16 16:51:39.051871] INFO: derived_feature_extractor: /y_2017, 443448

[2023-02-16 16:51:39.168078] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[1.587455s].

[2023-02-16 16:51:39.176577] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2023-02-16 16:51:39.183197] INFO: moduleinvoker: 命中缓存

[2023-02-16 16:51:39.184430] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[0.007859s].

[2023-02-16 16:51:39.192610] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2023-02-16 16:51:39.202291] INFO: moduleinvoker: 命中缓存

[2023-02-16 16:51:39.203461] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[0.010862s].

[2023-02-16 16:51:39.210534] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2023-02-16 16:51:40.646077] INFO: derived_feature_extractor: 提取完成 return_5_day = (shift(close_0, -5)-shift(open_0, -1))/shift(open_0, -1), 0.295s

[2023-02-16 16:51:40.649084] INFO: derived_feature_extractor: 提取完成 return_0/return_4, 0.002s

[2023-02-16 16:51:41.253161] INFO: derived_feature_extractor: /y_2017, 486868

[2023-02-16 16:51:41.972250] INFO: derived_feature_extractor: /y_2018, 513321

[2023-02-16 16:51:42.326803] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[3.116254s].

[2023-02-16 16:51:42.336417] INFO: moduleinvoker: join.v3 开始运行..

[2023-02-16 16:51:43.468838] INFO: join: /y_2016, 行数=0/165781, 耗时=0.289651s

[2023-02-16 16:51:44.181287] INFO: join: /y_2017, 行数=443448/443448, 耗时=0.709623s

[2023-02-16 16:51:44.258438] INFO: join: 最终行数: 443448

[2023-02-16 16:51:44.264828] INFO: moduleinvoker: join.v3 运行完成[1.928429s].

[2023-02-16 16:51:44.273438] INFO: moduleinvoker: join.v3 开始运行..

[2023-02-16 16:51:45.561684] INFO: join: /y_2017, 行数=299785/486868, 耗时=0.71511s

[2023-02-16 16:51:46.875861] INFO: join: /y_2018, 行数=513321/513321, 耗时=0.789764s

[2023-02-16 16:51:46.951103] INFO: join: 最终行数: 813106

[2023-02-16 16:51:46.958015] INFO: moduleinvoker: join.v3 运行完成[2.684565s].

[2023-02-16 16:51:46.966211] INFO: moduleinvoker: join.v3 开始运行..

[2023-02-16 16:51:48.055153] INFO: join: /y_2016, 行数=0/0, 耗时=0.075986s

[2023-02-16 16:51:48.917889] INFO: join: /y_2017, 行数=423687/443448, 耗时=0.861355s

[2023-02-16 16:51:48.980079] INFO: join: 最终行数: 423687

[2023-02-16 16:51:48.986913] INFO: moduleinvoker: join.v3 运行完成[2.020701s].

[2023-02-16 16:51:49.001679] INFO: moduleinvoker: filtet_st_stock.v7 开始运行..

[2023-02-16 16:51:51.490738] INFO: moduleinvoker: filtet_st_stock.v7 运行完成[2.489047s].

[2023-02-16 16:51:51.502527] INFO: moduleinvoker: filtet_st_stock.v7 开始运行..

[2023-02-16 16:51:53.287799] INFO: moduleinvoker: filtet_st_stock.v7 运行完成[1.785269s].

[2023-02-16 16:51:53.299979] INFO: moduleinvoker: filter_stockmarket.v2 开始运行..

[2023-02-16 16:51:55.027295] INFO: moduleinvoker: filter_stockmarket.v2 运行完成[1.727303s].

[2023-02-16 16:51:55.040274] INFO: moduleinvoker: filter_stockmarket.v2 开始运行..

[2023-02-16 16:51:56.337065] INFO: moduleinvoker: filter_stockmarket.v2 运行完成[1.296793s].

[2023-02-16 16:51:56.347891] INFO: moduleinvoker: filter_stockmarket.v2 开始运行..

[2023-02-16 16:51:57.590536] INFO: moduleinvoker: filter_stockmarket.v2 运行完成[1.242639s].

[2023-02-16 16:51:57.600896] INFO: moduleinvoker: filter_stockmarket.v2 开始运行..

[2023-02-16 16:51:58.684839] INFO: moduleinvoker: filter_stockmarket.v2 运行完成[1.083942s].

[2023-02-16 16:51:58.698160] INFO: moduleinvoker: aa.v5 开始运行..

[2023-02-16 16:52:02.672611] INFO: moduleinvoker: aa.v5 运行完成[3.974441s].

[2023-02-16 16:52:02.683185] INFO: moduleinvoker: aa.v5 开始运行..

[2023-02-16 16:52:05.252546] INFO: moduleinvoker: aa.v5 运行完成[2.56936s].

[2023-02-16 16:52:05.260791] INFO: moduleinvoker: dropnan.v1 开始运行..

[2023-02-16 16:52:06.376815] INFO: dropnan: /data, 604783/618306

[2023-02-16 16:52:06.465721] INFO: dropnan: 行数: 604783/618306

[2023-02-16 16:52:06.472763] INFO: moduleinvoker: dropnan.v1 运行完成[1.211966s].

[2023-02-16 16:52:06.481676] INFO: moduleinvoker: dropnan.v1 开始运行..

[2023-02-16 16:52:07.472441] INFO: dropnan: /data, 323470/323482

[2023-02-16 16:52:07.519856] INFO: dropnan: 行数: 323470/323482

[2023-02-16 16:52:07.526572] INFO: moduleinvoker: dropnan.v1 运行完成[1.044893s].

[2023-02-16 16:52:07.533799] INFO: moduleinvoker: stock_ranker_train.v6 开始运行..

[2023-02-16 16:52:08.309094] INFO: StockRanker: 特征预处理 ..

[2023-02-16 16:52:08.435134] INFO: StockRanker: prepare data: training ..

[2023-02-16 16:52:08.496912] INFO: StockRanker: sort ..

[2023-02-16 16:52:10.926058] INFO: StockRanker训练: 3155038e 准备训练: 323470 行数

[2023-02-16 16:52:10.927665] INFO: StockRanker训练: AI模型训练，将在323470*2=64.69万数据上对模型训练进行2轮迭代训练。预计将需要1～2分钟。请耐心等待。

[2023-02-16 16:52:11.099971] INFO: StockRanker训练: 正在训练 ..

[2023-02-16 16:52:11.165947] INFO: StockRanker训练: 任务状态: Pending

[2023-02-16 16:52:21.217489] INFO: StockRanker训练: 任务状态: Running

[2023-02-16 16:53:21.498311] INFO: StockRanker训练: 00:01:04.1078503, finished iteration 1

[2023-02-16 16:53:21.499717] INFO: StockRanker训练: 00:01:07.0680610, finished iteration 2

[2023-02-16 16:53:31.544605] INFO: StockRanker训练: 任务状态: Succeeded

[2023-02-16 16:53:31.733446] INFO: moduleinvoker: stock_ranker_train.v6 运行完成[84.199618s].

[2023-02-16 16:53:31.741826] INFO: moduleinvoker: stock_ranker_predict.v5 开始运行..

[2023-02-16 16:53:32.489308] INFO: StockRanker预测: /data ..

[2023-02-16 16:53:33.064729] INFO: moduleinvoker: stock_ranker_predict.v5 运行完成[1.322893s].

[2023-02-16 16:53:33.072144] INFO: moduleinvoker: join.v3 开始运行..

[2023-02-16 16:53:35.035271] INFO: join: /data, 行数=604783/604783, 耗时=1.276441s

[2023-02-16 16:53:35.110573] INFO: join: 最终行数: 604783

[2023-02-16 16:53:35.116780] INFO: moduleinvoker: join.v3 运行完成[2.044628s].

[2023-02-16 16:53:35.126942] INFO: moduleinvoker: score_group_filter.v3 开始运行..

[2023-02-16 16:53:35.131299] ERROR: moduleinvoker: module name: score_group_filter, module version: v3, trackeback: TypeError: bigquant_run() got an unexpected keyword argument 'number_of_trees'

[2023-02-16 16:53:35.134075] ERROR: moduleinvoker: module name: hyper_rolling_train, module version: v1, trackeback: TypeError: bigquant_run() got an unexpected keyword argument 'number_of_trees'

========================= 全部滚动的时间区间为: [{'train_start_date': '2017-01-03', 'train_end_date': '2017-08-14', 'test_start_date': '2017-08-15', 'test_end_date': '2018-08-21'}, {'train_start_date': '2017-08-15', 'train_end_date': '2018-08-21', 'test_start_date': '2018-08-22', 'test_end_date': '2019-08-30'}, {'train_start_date': '2018-08-22', 'train_end_date': '2019-08-30', 'test_start_date': '2019-09-02', 'test_end_date': '2020-09-10'}, {'train_start_date': '2019-09-02', 'train_end_date': '2020-09-10', 'test_start_date': '2020-09-11', 'test_end_date': '2021-09-22'}, {'train_start_date': '2020-09-11', 'train_end_date': '2021-09-22', 'test_start_date': '2021-09-23', 'test_end_date': '2022-10-10'}, {'train_start_date': '2021-09-23', 'train_end_date': '2022-10-10', 'test_start_date': '2022-10-11', 'test_end_date': '2023-02-10'}]
本次滚动回测的参数为: 10 0.1 {'train_start_date': '2017-01-03', 'train_end_date': '2017-08-14', 'test_start_date': '2017-08-15', 'test_end_date': '2018-08-21'} {'m6.number_of_trees': 10, 'm6.learning_rate': 0.1, 'm6.plot_charts': False, 'm19.plot_charts': False, 'm29.start_date': '2017-01-03', 'm29.end_date': '2017-08-14', 'm9.start_date': '2017-08-15', 'm9.end_date': '2018-08-21'}

设置评估测试数据集，查看训练曲线
[视频教程]StockRanker训练曲线

bigcharts-data-start/{"__type":"tabs","__id":"bigchart-7db079a041e64a939072b6953a391800"}/bigcharts-data-end

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-40-3b39bf1bc069> in <module>
    546 
    547 
--> 548 m28 = M.hyper_rolling_train.v1(
    549     run=m28_run_bigquant_run,
    550     run_now=True,

<ipython-input-40-3b39bf1bc069> in m28_run_bigquant_run(bq_graph, inputs, trading_days_market, train_instruments_mid, test_instruments_mid, predict_mid, trade_mid, model_train_mid, start_date, end_date, train_update_days, train_update_days_for_live, train_data_min_days, train_data_max_days, rolling_count_for_live)
    499                 print('本次滚动回测的参数为:', nt, lr, rolling, parameters)
    500 
--> 501                 output = {'参数详情':(nt,lr), 'graph_data':g.run(parameters)} # 将参数和图的结果保存下来
    502                 results.append(output)
    503 

TypeError: bigquant_run() got an unexpected keyword argument 'number_of_trees'