labels_2.py

def get_t_events(raw_price, threshold):
    """
    :param raw_price: (series) of close prices.
    :param threshold: (float) when the abs(change) is larger than the threshold, the
    function captures it as an event.
    :return: (datetime index vector) vector of datetimes when the events occurred. This is used later to sample.
    """
    print('Applying Symmetric CUSUM filter.')

    t_events = []
    s_pos = 0
    s_neg = 0

    # log returns
    diff = np.log(raw_price).diff().dropna()

    # Get event time stamps for the entire series
    for i in tqdm(diff.index[1:]):
        pos = float(s_pos + diff.loc[i])
        neg = float(s_neg + diff.loc[i])
        s_pos = max(0.0, pos)
        s_neg = min(0.0, neg)

        if s_neg < -threshold:
            s_neg = 0
            t_events.append(i)

        elif s_pos > threshold:
            s_pos = 0
            t_events.append(i)

    event_timestamps = pd.DatetimeIndex(t_events)
    return event_timestamps