environment.py 文件源码

def filterout_outliers(l_data, l_date):
    '''
    Return a list with data filtered from outliers

    :param l_data: list. data to be analyzed
    '''
    # === old code to filter outliers =======
    # Q3 = np.percentile(l_data, 98)
    # Q1 = np.percentile(l_data, 2)
    # step = (Q3 - Q1) * 1.5
    # step = max(3000., step)
    # na_val = np.array(l_data)
    # na_val = na_val[(na_val >= Q1 - step) & (na_val <= Q3 + step)]
    # return na_val
    # =======================================
    # group by minute
    df_filter = pd.Series(np.array(l_date)/60).astype(int)
    l_filter = list((df_filter != df_filter.shift()).values)
    l_filter[0] = True
    l_filter[-1] = True
    return np.array(pd.Series(l_data)[l_filter].values)