示例#1
0
def main():
    from data.const import get_trade_calendar
    from datetime import time

    trade_cal = get_trade_calendar()

    def transform_tick(tick):
        ret = None
        if tick in ['000001', '000300']:
            ret = ''.join((tick, '_SH_IX'))
        else:
            ret = ''.join((tick, '_SZ_IX'))
        return ret

    today = str(datetime.now().date())
    today_idx = trade_cal.index.searchsorted(today)
    yesterday = str(trade_cal.index[today_idx - 1].date())
    tomorrow = str(trade_cal.index[today_idx + 1].date())

    from_date, to_date = None, None
    if datetime.now().time() > time(15, 10, 0, 0):  # 过了15:10
        from_date, to_date = today, tomorrow
    else:
        from_date, to_date = yesterday, today

    index_list = ['000001', '000300', '399001', '399005', '399006']

    get_config_func = make_config(CONFIG_FILE)
    ret = predict(ticks=index_list, look_back=519, index=True,
                  get_config_func=get_config_func)
    ret = ret.sort_values(by='dt')
    print(ret)
    engines = [engine_company_local,
               engine_company_outer]
    for engine in engines:
        Base.metadata.bind = engine
        # Base.metadata.create_all(engine)
        Session.configure(bind=engine)
        session = Session()
        IndexPrediction().to_sql(session, ret)

    hist_rec = pd.read_sql_table('index_prediction', engine_company_outer)

    win_ratio = (hist_rec.dropna()
                 .query('dt>"2014-01-01"')
                 .groupby('secu')
                 .apply(lambda df: np.sum(
                     np.isclose(df.pred, df.fact)) * 1.0 / len(df))
                 .to_frame()
                 .rename(columns={0: 'accuracy'})
                 .reset_index()
                 )

    tomorrow_pred = ret.query("""dt == '{}'""".format(to_date))

    to_seven_liu = pd.merge(tomorrow_pred, win_ratio, on='secu')
    logger.info(u'明日预测结果:\n{}'.format(to_seven_liu))
示例#2
0
def main():
    from data.const import get_trade_calendar
    from datetime import time

    trade_cal = get_trade_calendar()

    def transform_tick(tick):
        ret = None
        if tick in ['000001', '000300']:
            ret = ''.join((tick, '_SH_IX'))
        else:
            ret = ''.join((tick, '_SZ_IX'))
        return ret

    today = str(datetime.now().date())
    today_idx = trade_cal.index.searchsorted(today)
    yesterday = str(trade_cal.index[today_idx - 1].date())
    tomorrow = str(trade_cal.index[today_idx + 1].date())

    from_date, to_date = None, None
    if datetime.now().time() > time(15, 10, 0, 0):  # 过了15:10
        from_date, to_date = today, tomorrow
    else:
        from_date, to_date = yesterday, today

    # tomorrow = str(trade_cal.index[today_idx + 1].date())

    index_list = ['000001', '000300', '399001', '399005', '399006']
    # index_list = ['000001']
    # today = str(datetime.now().date())

    ret = []
    for idx in index_list:
        ret.append(predict_history(idx,
                                   from_date=from_date,
                                   to_date=to_date,
                                   is_index=True,
                                   data_source=['mysql', 'local'],
                                   config_params=config_params))
    df = pd.concat(ret)
    # df.index = df.index.format()
    df.reset_index(inplace=True)
    cols_to_rename = {'dates': 'dt', 'index': 'dt', 'prediction': 'pred', 'facts': 'fact', 'tick': 'secu'}
    df.rename(columns=cols_to_rename, inplace=True)
    logger.debug('df in predict local is \n{}'.format(df))
    df.dt = df.dt.apply(lambda x: str(x.date()))
    df.secu = df.secu.apply(transform_tick)

    df = df.where(pd.notnull(df), None)
    df = df.loc[:, ['dt', 'secu', 'pred', 'fact']]
    print(df.sort_values(by=['dt', 'secu']))
    # Session.configure(bind=engine_phil_machine_test)
    engines = [
        # engine_phil_machine_test,
        engine_phil_machine,
        # engine_company_local,
        # engine_company_outer,
    ]
    for engine in engines:
        Base.metadata.bind = engine
        Base.metadata.create_all(engine)
        Session.configure(bind=engine)
        session = Session()
        IndexPrediction().to_sql(session, df)