复制链接

克隆策略

In [14]:

m38.data.read_df()

Out[14]:

	date	instrument	normalize(return_0)	normalize(mf_net_amount_s_0)	m:amount	m:close	m:low	m:high	m:open	label
0	2022-09-01	000001.SZA	-0.285774	0.609585	1.092666e+09	1436.735596	1433.317505	1457.244141	1441.293091	-1.069725
1	2022-09-01	000002.SZA	0.674134	-0.440758	2.216182e+09	2910.356689	2846.411621	2972.573242	2894.802490	-0.059062
4	2022-09-01	000006.SZA	1.126484	-0.135633	1.090169e+08	167.829453	163.935486	170.555222	165.493073	2.097905
5	2022-09-01	000007.SZA	2.171985	-0.245828	1.954950e+08	70.575661	66.516731	70.741325	66.599564	1.959467
6	2022-09-01	000008.SZA	-0.345454	-0.357629	4.120158e+07	53.331738	53.331738	54.452152	53.779903	1.882845
...	...	...	...	...	...	...	...	...	...	...
75163	2023-01-30	605580.SHA	-0.079891	-0.210726	1.619372e+07	13.014465	12.931305	13.056045	12.931305	0.877905
75164	2023-01-30	605588.SHA	-0.358629	-0.092787	8.011155e+06	31.306278	31.276157	31.617535	31.446846	2.969187
75165	2023-01-30	605589.SHA	1.210585	0.479374	2.413766e+08	23.019468	22.389078	23.619358	22.389078	3.377485
75166	2023-01-30	605598.SHA	-0.938862	0.010872	6.262655e+07	27.966263	27.765787	28.497520	28.287022	5.839288
75167	2023-01-30	605599.SHA	-0.676580	-0.252099	6.284808e+07	10.442697	10.318870	10.659392	10.504610	1.198798

406322 rows × 10 columns

In [13]:

# 本代码由可视化策略环境自动生成 2023年5月26日 10:48
# 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。


import talib as ta
def wma(df,x):
    return 1/(1+np.exp(-1*x))
def hl(df,x,time):
    return ta.LINEARREG_ANGLE(x,time)
m16_user_functions_bigquant_run = {
    'ht': wma,
    'angle': hl
    
}
import talib as ta
def wma(df,x):
    return 1/(1+np.exp(-1*x))
def hl(df,x,time):
    return ta.LINEARREG_ANGLE(x,time)
m17_user_functions_bigquant_run = {
    'ht': wma,
    'angle': hl
    
}
# 回测引擎：初始化函数，只执行一次
def m14_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)])
    #改为等权重配置
    context.stock_weights = [1 / stock_count 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 = 1
    #context.max_cash_per_instrument = 1
    context.options['hold_days'] = 1

# 回测引擎：每日数据处理函数，每天执行一次
def m14_handle_data_bigquant_run(context, data):
    # 按日期过滤得到今日的预测数据
    today = data.current_dt.strftime('%Y-%m-%d')
    ranker_prediction = context.ranker_prediction[
    context.ranker_prediction.date == today]
    sell = 1000
    buy = 0
    # 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_for_buy = min(context.portfolio.cash, context.portfolio.portfolio_value*0.5)
    
    positions = {e.symbol: p.amount * p.last_sale_price
                 for e, p in context.perf_tracker.position_tracker.positions.items()}
   
    
    # 2. 生成买入订单：按StockRanker预测的排序，买入前面的stock_count只股票
    
    rank_buy = ranker_prediction[ranker_prediction.score>buy]
    buy_instruments1 = list(rank_buy.instrument)
    buy_instruments = buy_instruments1[:np.where(len(buy_instruments1)>1,1,len(buy_instruments1))]
    
    
    # 2. ST股和退市股的卖出
    stock_sold = [] # 记录卖出的股票，防止多次卖出出现空单
    name_df = context.name_df
    name_today = name_df[name_df.date==today]
    #-------------------------- START: ST和退市股卖出 ---------------------  
    st_stock_list = []
    for instrument in positions.keys():
        try:
            instrument_name = name_today[name_today.instrument==instrument]['name'].values[0]
            # 如果股票状态变为了st或者退市 则卖出
            if 'ST' in instrument_name or '退' in instrument_name or '*' in instrument_name:
                
                if instrument in stock_sold:
                    continue
                if data.can_trade(context.symbol(instrument)):
                    context.order_target(context.symbol(instrument), 0)
                    st_stock_list.append(instrument)
                    cash_for_sell -= positions[instrument]
        except:
            continue
    if st_stock_list!=[]:
        #print(today,'持仓出现st股/退市股',st_stock_list,'进行卖出处理')    
        stock_sold += st_stock_list

    #-------------------------- END: ST和退市股卖出 --------------------- 
    
    equities = {e.symbol: e for e, p in context.perf_tracker.position_tracker.positions.items()}
    # 3. 生成卖出订单：hold_days天之后才开始卖出；对持仓的股票，按StockRanker预测的排序末位淘汰

    sell_instruments_back = equities
    sell_instruments = list(set(sell_instruments_back) - set(buy_instruments))
        #sell_instruments = list(set(sell_instruments_back) - set(buy_instruments))
    for instrument in sell_instruments:
        context.order_target(context.symbol(instrument), 0)
    # 4. 生成买入订单：按StockRanker预测的排序，买入前面的stock_count只股票
    
    buy_scores1 = list(rank_buy.score)
    buy_scores = buy_scores1[:len(buy_instruments)]
    buy_cash_weights = buy_scores/np.sum(buy_scores)
    max_cash_per_instrument = context.portfolio.portfolio_value * context.max_cash_per_instrument
    if any(buy_instruments):
        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)
            today_price = data.current(symbol(instrument), ['close','adjust_factor'])
            buy_price =today_price['close'] / today_price['adjust_factor']
            buy_amount = int((cash/(buy_price*100)))
            if buy_amount > 0: 
                context.order_lots(symbol(instrument),buy_amount)
# 回测引擎：准备数据，只执行一次
def m14_prepare_bigquant_run(context):
    # 获取股票名称 用于过滤st和退市股
    
   #卖出ST股
    context.name_df = DataSource('instruments_CN_STOCK_A').read()
   


m1 = M.instruments.v2(
    start_date='2022-09-01',
    end_date='2023-02-01',
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m10 = M.advanced_auto_labeler.v2(
    instruments=m1.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/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>`_

(((shift(close, -2) / shift(open, -1)-1)/2+min(shift(close,-1)/shift(open,-1)-1,shift(close,-2)/shift(close,-1)-1))*100)

# 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, 60)

# 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,
    user_functions={}
)

m3 = M.input_features.v1(
    features="""normalize(return_0)
normalize(mf_net_amount_s_0)



        
        """
)

m15 = M.general_feature_extractor.v7(
    instruments=m1.data,
    features=m3.data,
    start_date='',
    end_date='',
    before_start_days=100
)

m16 = M.derived_feature_extractor.v3(
    input_data=m15.data,
    features=m3.data,
    date_col='date',
    instrument_col='instrument',
    drop_na=False,
    remove_extra_columns=True,
    user_functions=m16_user_functions_bigquant_run
)

m7 = M.join.v3(
    data1=m10.data,
    data2=m16.data,
    on='date,instrument',
    how='inner',
    sort=True
)

m6 = M.dropnan.v2(
    input_data=m7.data
)

m5 = M.chinaa_stock_filter.v1(
    input_data=m6.data,
    index_constituent_cond=['全部'],
    board_cond=['上证主板', '深证主板', '创业板'],
    industry_cond=['全部'],
    st_cond=['正常'],
    delist_cond=['非退市'],
    output_left_data=False
)

m38 = M.winsorize.v7(
    input_data=m5.data,
    features=m3.data,
    columns_input='',
    function_name='MAD',
    group='date'
)

m8 = M.stock_ranker_train.v6(
    training_ds=m38.data,
    features=m3.data,
    learning_algorithm='回归',
    number_of_leaves=30,
    minimum_docs_per_leaf=1000,
    number_of_trees=20,
    learning_rate=0.1,
    max_bins=1023,
    feature_fraction=1,
    data_row_fraction=1,
    plot_charts=True,
    ndcg_discount_base=1,
    m_lazy_run=False
)

m4 = M.instruments.v2(
    start_date='2023-02-02',
    end_date='2023-05-24',
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m13 = M.general_feature_extractor.v7(
    instruments=m4.data,
    features=m3.data,
    start_date='',
    end_date='',
    before_start_days=100
)

m17 = M.derived_feature_extractor.v3(
    input_data=m13.data,
    features=m3.data,
    date_col='date',
    instrument_col='instrument',
    drop_na=False,
    remove_extra_columns=True,
    user_functions=m17_user_functions_bigquant_run
)

m18 = M.dropnan.v2(
    input_data=m17.data
)

m23 = M.chinaa_stock_filter.v1(
    input_data=m18.data,
    index_constituent_cond=['全部'],
    board_cond=['上证主板', '深证主板', '创业板'],
    industry_cond=['全部'],
    st_cond=['正常'],
    delist_cond=['非退市'],
    output_left_data=False
)

m22 = M.winsorize.v7(
    input_data=m23.data,
    features=m3.data,
    columns_input='',
    function_name='MAD',
    group='date'
)

m9 = M.stock_ranker_predict.v5(
    model=m8.model,
    data=m22.data,
    m_lazy_run=False
)

m14 = M.trade.v4(
    instruments=m4.data,
    options_data=m9.predictions,
    start_date='',
    end_date='',
    initialize=m14_initialize_bigquant_run,
    handle_data=m14_handle_data_bigquant_run,
    prepare=m14_prepare_bigquant_run,
    volume_limit=0,
    order_price_field_buy='open',
    order_price_field_sell='close',
    capital_base=100000,
    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'
)

[2023-05-26 10:40:27.416331] INFO: moduleinvoker: instruments.v2 开始运行..

[2023-05-26 10:40:27.424065] INFO: moduleinvoker: 命中缓存

[2023-05-26 10:40:27.426602] INFO: moduleinvoker: instruments.v2 运行完成[0.010271s].

[2023-05-26 10:40:27.436745] INFO: moduleinvoker: advanced_auto_labeler.v2 开始运行..

[2023-05-26 10:40:27.444186] INFO: moduleinvoker: 命中缓存

[2023-05-26 10:40:27.445874] INFO: moduleinvoker: advanced_auto_labeler.v2 运行完成[0.009134s].

[2023-05-26 10:40:27.450161] INFO: moduleinvoker: input_features.v1 开始运行..

[2023-05-26 10:40:27.463768] INFO: moduleinvoker: 命中缓存

[2023-05-26 10:40:27.465606] INFO: moduleinvoker: input_features.v1 运行完成[0.015451s].

[2023-05-26 10:40:27.479060] INFO: moduleinvoker: general_feature_extractor.v7 开始运行..

[2023-05-26 10:40:27.484648] INFO: moduleinvoker: 命中缓存

[2023-05-26 10:40:27.485978] INFO: moduleinvoker: general_feature_extractor.v7 运行完成[0.006932s].

[2023-05-26 10:40:27.493173] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2023-05-26 10:40:27.498922] INFO: moduleinvoker: 命中缓存

[2023-05-26 10:40:27.500258] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[0.00708s].

[2023-05-26 10:40:27.506841] INFO: moduleinvoker: join.v3 开始运行..

[2023-05-26 10:40:27.511609] INFO: moduleinvoker: 命中缓存

[2023-05-26 10:40:27.512722] INFO: moduleinvoker: join.v3 运行完成[0.005879s].

[2023-05-26 10:40:27.519345] INFO: moduleinvoker: dropnan.v2 开始运行..

[2023-05-26 10:40:27.524089] INFO: moduleinvoker: 命中缓存

[2023-05-26 10:40:27.525230] INFO: moduleinvoker: dropnan.v2 运行完成[0.005883s].

[2023-05-26 10:40:27.531840] INFO: moduleinvoker: chinaa_stock_filter.v1 开始运行..

[2023-05-26 10:40:27.537261] INFO: moduleinvoker: 命中缓存

[2023-05-26 10:40:27.538416] INFO: moduleinvoker: chinaa_stock_filter.v1 运行完成[0.006577s].

[2023-05-26 10:40:27.542647] INFO: moduleinvoker: winsorize.v7 开始运行..

[2023-05-26 10:40:27.555526] INFO: moduleinvoker: 命中缓存

[2023-05-26 10:40:27.556819] INFO: moduleinvoker: winsorize.v7 运行完成[0.014169s].

[2023-05-26 10:40:27.563166] INFO: moduleinvoker: stock_ranker_train.v6 开始运行..

[2023-05-26 10:40:27.585325] INFO: StockRanker训练: ac66692a 准备训练: 406322 行数

[2023-05-26 10:40:27.586586] INFO: StockRanker训练: AI模型训练，将在406322*2=81.26万数据上对模型训练进行20轮迭代训练。预计将需要1～2分钟。请耐心等待。

[2023-05-26 10:40:27.866025] INFO: StockRanker训练: 正在训练 ..

[2023-05-26 10:40:27.917618] INFO: StockRanker训练: 任务状态: Pending

[2023-05-26 10:40:37.957766] INFO: StockRanker训练: 任务状态: Running

[2023-05-26 10:41:38.224255] INFO: StockRanker训练: 任务状态: Succeeded

[2023-05-26 10:41:38.234834] ERROR: moduleinvoker: module name: stock_ranker_train, module version: v6, trackeback: Exception: 模型训练失败：可能导致错误的原因是训练数据问题，请检查训练数据, err_code=1 (ac66692afb6e11ed90915a6d36c7f5bf)

---------------------------------------------------------------------------
Exception                                 Traceback (most recent call last)
<ipython-input-13-90ed6cb825ca> in <module>
    221 )
    222 
--> 223 m8 = M.stock_ranker_train.v6(
    224     training_ds=m38.data,
    225     features=m3.data,

Exception: 模型训练失败：可能导致错误的原因是训练数据问题，请检查训练数据, err_code=1 (ac66692afb6e11ed90915a6d36c7f5bf)