vn.py量化社区
By Traders, For Traders.
Member
加入于:
帖子: 58
声望: 9

之前2月25日,上海期货交易所进行测试,在周六进行行情广播,我的datarecording一直在跑;然后就发现读了不少脏数据。
vnpy自带的行情清理功能较为简单,只是在清除非交易时段,没有考虑周六日;而且只是笼统给了一个最大时间交易范围,像股指期货没有夜盘,螺纹钢晚上11点就结束,但是默认只是结束在凌晨两点半这个最大交易时间。 所以写了一个方法,按照不同品种,更细致的清理。

可以直接把这个方法插入\DataRecording\runDataCleaning.py, 然后替代原来方法。也可以自己另外调用。

# ----------------------------------------------------------------------
def cleanDataAdv(dbName, collectionName, start):
    """清洗数据"""
    #新的静态数据
    # 这里以商品期货为例
    MORNING_START = time(9, 0)
    MORNING_REST = time(10, 15)
    MORNING_RESTART = time(10, 30)
    MORNING_END = time(11, 30)
    AFTERNOON_START = time(13, 30)
    AFTERNOON_END = time(15, 0)
    NIGHT_START = time(21, 0)
    NIGHT_END = time(2, 30)
    #股指期货
    STOCK_FUTURE = ["IC", "IF", "IH"]
    MORNING_START_STOCK = time(9, 30)
    AFTERNOON_START_STOCK = time(13,0)
    AFTERNOON_END_STOCK = time(15, 0)
    #晚上11点结束交易,不全,请自行维护
    PM11CLOSE_FUTURE = ['rb','ru','bu','hc','sp']
    NIGHT_END_11 = time(23, 00)
    #晚上11点半结束交易,不全,请自行维护,大连只有一位标志,所以带1
    PM1130CLOSE_FUTURE = ['FG','MA','SR','TA','RM','OI','CF','CY','ZC','i1','j1','m1','p1','y1']
    NIGHT_END_1130 = time(23, 30)
    #凌晨1点半结束交易,不全,请自行维护
    AM1CLOSE_FUTURE = ['cu','pd','al','zn']
    NIGHT_END_AM1 = time(1, 00)
    print(u'\n清洗数据库:%s, 集合:%s, 起始日:%s' % (dbName, collectionName, start))
    mc = MongoClient('localhost', 27017)  # 创建MongoClient
    cl = mc[dbName][collectionName]  # 获取数据集合
    d = {'datetime': {'$gte': start}}  # 只过滤从start开始的数据
    cx = cl.find(d)  # 获取数据指针
    for data in cx:
    # 获取时间戳对象
        dt = data['datetime'].time()
        # 默认需要清洗
        cleanRequired = True
        ####如果是股指期货,这没有上午休息和夜盘,9点半到11点半,下午1点到下午三点,周六日无行情
        if collectionName[:2] in STOCK_FUTURE:
            if data['datetime'].weekday() is not (5 or 6):
                if ((MORNING_START_STOCK <= dt < MORNING_END) or
                        (AFTERNOON_START_STOCK <= dt < AFTERNOON_END_STOCK)):
                    cleanRequired = False
        ####如果是11点结束,则周六日无行情
        elif collectionName[:2] in PM11CLOSE_FUTURE:
            if data['datetime'].weekday() is not (5 or 6):
                if ((MORNING_START <= dt < MORNING_REST) or
                        (MORNING_RESTART <= dt < MORNING_END) or
                        (AFTERNOON_START <= dt < AFTERNOON_END) or
                        ( NIGHT_START <= dt <NIGHT_END_11)):
                    cleanRequired = False
        ####如果是11点半结束,则周六日无行情
        elif collectionName[:2] in PM1130CLOSE_FUTURE:
            if data['datetime'].weekday() is not (5 or 6):
                if ((MORNING_START <= dt < MORNING_REST) or
                        (MORNING_RESTART <= dt < MORNING_END) or
                        (AFTERNOON_START <= dt < AFTERNOON_END) or
                        (NIGHT_START <= dt < NIGHT_END_1130)):
                    cleanRequired = False
        ####如果是1点结束,
        elif collectionName[:2] in AM1CLOSE_FUTURE:
            # 如果在交易事件内,则为有效数据,无需清洗
            if data['datetime'].weekday() is not 6:
                if ((MORNING_START <= dt < MORNING_REST) or
                        (MORNING_RESTART <= dt < MORNING_END) or
                        (AFTERNOON_START <= dt < AFTERNOON_END) or
                        (dt >= NIGHT_START) or
                        (dt < NIGHT_END_AM1)):
                    cleanRequired = False
        else:
            # 如果在交易事件内,则为有效数据,无需清洗
            if data['datetime'].weekday() is not 6:
                if ((MORNING_START <= dt < MORNING_REST) or
                    (MORNING_RESTART <= dt < MORNING_END) or
                    (AFTERNOON_START <= dt < AFTERNOON_END) or
                    (dt >= NIGHT_START) or
                    (dt < NIGHT_END)):
                    cleanRequired = False
            # 如果需要清洗
        if cleanRequired:
            print(u'删除无效数据,时间戳:%s' % data['datetime'])
            cl.delete_one(data)
    print(u'清洗完成,数据库:%s, 集合:%s' % (dbName, collectionName))
Member
avatar
加入于:
帖子: 3
声望: 0

张老师能把vnpy2.0版解析一下吗?最好用思维导图xmind,谢谢

Member
加入于:
帖子: 58
声望: 9

wrote:

张老师能把vnpy2.0版解析一下吗?最好用思维导图xmind,谢谢

免贵,看了下2.0,改变地方非常多,还在学习中。vnpy 2.0可以更好和深度学习框架集成,很有搞头

Member
avatar
加入于:
帖子: 3
声望: 0

张老师,方便可以加下我的QQ交流:1687959304

Member
avatar
加入于:
帖子: 69
声望: 1

感谢张老师

从你的博客里面我收获良多。

Member
avatar
加入于:
帖子: 16
声望: 0

注释详细,可读性强

Member
avatar
加入于:
帖子: 1
声望: 0

张老师请问,每个品种非交易时间不要接收行情,怎么实现,请指导一下

© 2015-2019 上海韦纳软件科技有限公司
备案服务号:沪ICP备18006526号-3