概

这个笔记本提供了创建分层预测管道的逐步指南。

在该管道中，我们将使用 HierarchicalForecast 和 StatsForecast 核心类，来创建基本预测、进行协调和评估。

我们将使用 TourismL 数据集，该数据集总结了大型澳大利亚国家游客调查。

大纲 1. 安装包 2. 准备 TourismL 数据集 - 读取和聚合 - StatsForecast 的基本预测 3. 协调 4. 评估

在 Colab 中打开

1. 安装 HierarchicalForecast

我们假设您已经安装了 StatsForecast 和 HierarchicalForecast，如果没有，请查看本指南以获取有关如何安装 HierarchicalForecast 的说明。

# %%capture
# !pip 安装 hierarchicalforecast
# !pip 安装 -U numba statsforecast datasetsforecast
import os
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

from statsforecast.core import StatsForecast
from statsforecast.models import AutoARIMA, Naive

from hierarchicalforecast.core import HierarchicalReconciliation
from hierarchicalforecast.evaluation import HierarchicalEvaluation
from hierarchicalforecast.methods import BottomUp, TopDown, MinTrace, ERM

from hierarchicalforecast.utils import is_strictly_hierarchical
from hierarchicalforecast.utils import HierarchicalPlot, CodeTimer
from hierarchicalforecast.evaluation import scaled_crps, msse, energy_score

from datasetsforecast.hierarchical import HierarchicalData, HierarchicalInfo
/Users/cchallu/opt/anaconda3/envs/hierarchicalforecast/lib/python3.10/site-packages/statsforecast/core.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
  from tqdm.autonotebook import tqdm

2. 准备旅游L数据集

2.1 读取层次数据集

# ['Labour', 'Traffic', 'TourismSmall', 'TourismLarge', 'Wiki2']
dataset = 'TourismSmall' # 'TourismLarge'
verbose = True
intervals_method = 'bootstrap'
LEVEL = np.arange(0, 100, 2)
qs = [[50-lv/2, 50+lv/2] for lv in LEVEL]
QUANTILES = np.sort(np.concatenate(qs)/100)
with CodeTimer('Read and Parse data   ', verbose):
    print(f'{dataset}')
    if not os.path.exists('./data'):
        os.makedirs('./data')
    
    dataset_info = HierarchicalInfo[dataset]
    Y_df, S_df, tags = HierarchicalData.load(directory=f'./data/{dataset}', group=dataset)
    Y_df['ds'] = pd.to_datetime(Y_df['ds'])

    # 训练/测试分割
    horizon = dataset_info.horizon
    seasonality = dataset_info.seasonality
    Y_test_df = Y_df.groupby('unique_id').tail(horizon)
    Y_train_df = Y_df.drop(Y_test_df.index)
    Y_test_df = Y_test_df.set_index('unique_id')
    Y_train_df = Y_train_df.set_index('unique_id')
TourismSmall
100%|██████████| 1.30M/1.30M [00:00<00:00, 2.74MiB/s]
INFO:datasetsforecast.utils:Successfully downloaded datasets.zip, 1297279, bytes.
INFO:datasetsforecast.utils:Decompressing zip file...
INFO:datasetsforecast.utils:Successfully decompressed data/TourismSmall/hierarchical/datasets.zip
Code block 'Read and Parse data   ' took:   0.99873 seconds
dataset_info.seasonality
4
hplot = HierarchicalPlot(S=S_df, tags=tags)
hplot.plot_summing_matrix()

Y_train_df
ds y
unique_id
total 1998-03-31 84503
total 1998-06-30 65312
total 1998-09-30 72753
total 1998-12-31 70880
total 1999-03-31 86893
... ... ...
nt-oth-noncity 2003-12-31 132
nt-oth-noncity 2004-03-31 12
nt-oth-noncity 2004-06-30 40
nt-oth-noncity 2004-09-30 186
nt-oth-noncity 2004-12-31 144

2492 rows × 2 columns

2.2 StatsForecast的基础预测

该单元计算使用StatsForecast的AutoARIMA对Y_df中所有序列的基础预测Y_hat_df。此外，我们还获得了需要内插预测的算法的内插预测Y_fitted_df。

%%capture
with CodeTimer('Fit/Predict Model     ', verbose):
    # 阅读以避免不必要的AutoARIMA计算
    yhat_file = f'./data/{dataset}/Y_hat.csv'
    yfitted_file = f'./data/{dataset}/Y_fitted.csv'

    if os.path.exists(yhat_file):
        Y_hat_df = pd.read_csv(yhat_file)
        Y_fitted_df = pd.read_csv(yfitted_file)

        Y_hat_df = Y_hat_df.set_index('unique_id')
        Y_fitted_df = Y_fitted_df.set_index('unique_id')

    else:
        fcst = StatsForecast(
            df=Y_train_df, 
            models=[AutoARIMA(season_length=seasonality)],
            fallback_model=[Naive()],
            freq='M', 
            n_jobs=-1
        )
        Y_hat_df = fcst.forecast(h=horizon, fitted=True, level=LEVEL)
        Y_fitted_df = fcst.forecast_fitted_values()
        Y_hat_df.to_csv(yhat_file)
        Y_fitted_df.to_csv(yfitted_file)
Y_hat_df
ds AutoARIMA AutoARIMA-lo-98 AutoARIMA-lo-96 AutoARIMA-lo-94 AutoARIMA-lo-92 AutoARIMA-lo-90 AutoARIMA-lo-88 AutoARIMA-lo-86 AutoARIMA-lo-84 ... AutoARIMA-hi-80 AutoARIMA-hi-82 AutoARIMA-hi-84 AutoARIMA-hi-86 AutoARIMA-hi-88 AutoARIMA-hi-90 AutoARIMA-hi-92 AutoARIMA-hi-94 AutoARIMA-hi-96 AutoARIMA-hi-98
unique_id
bus 2005-01-31 9673.424805 7436.356445 7698.493652 7864.811523 7989.925781 8091.696289 8178.319336 8254.270508 8322.276367 ... 10905.793945 10962.725586 11024.573242 11092.579102 11168.530273 11255.153320 11356.923828 11482.038086 11648.356445 11910.493164
bus 2005-02-28 10393.900391 8156.831543 8418.968750 8585.287109 8710.401367 8812.171875 8898.794922 8974.746094 9042.751953 ... 11626.269531 11683.200195 11745.048828 11813.054688 11889.005859 11975.628906 12077.399414 12202.513672 12368.832031 12630.968750
bus 2005-03-31 12028.134766 9791.066406 10053.204102 10219.521484 10344.635742 10446.406250 10533.029297 10608.981445 10676.986328 ... 13260.503906 13317.435547 13379.283203 13447.289062 13523.240234 13609.863281 13711.633789 13836.748047 14003.066406 14265.203125
bus 2005-04-30 10995.679688 8758.610352 9020.748047 9187.065430 9312.179688 9413.951172 9500.574219 9576.525391 9644.531250 ... 12228.047852 12284.979492 12346.828125 12414.833008 12490.785156 12577.407227 12679.178711 12804.292969 12970.610352 13232.748047
bus 2005-05-31 9673.424805 7262.085449 7544.643555 7723.917480 7858.778320 7968.477539 8061.848633 8143.716797 8217.019531 ... 11001.796875 11063.164062 11129.830078 11203.132812 11285.000977 11378.372070 11488.071289 11622.932617 11802.206055 12084.764648
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
wa-vfr-noncity 2005-04-30 904.125549 463.371521 515.018616 547.787048 572.437439 592.488647 609.555359 624.519531 637.918213 ... 1146.930542 1158.147339 1170.332886 1183.731567 1198.695679 1215.762451 1235.813721 1260.464111 1293.232544 1344.879517
wa-vfr-noncity 2005-05-31 904.125549 457.607361 509.929901 543.126831 568.099670 588.413086 605.703003 620.862854 634.436707 ... 1150.105957 1161.469482 1173.814331 1187.388184 1202.548096 1219.838013 1240.151489 1265.124268 1298.321167 1350.643677
wa-vfr-noncity 2005-06-30 904.125549 451.916687 504.906036 538.526062 563.817139 584.389465 601.899719 617.252808 630.999634 ... 1153.240967 1164.749268 1177.251465 1190.998291 1206.351440 1223.861694 1244.433960 1269.724976 1303.345093 1356.334473
wa-vfr-noncity 2005-07-31 904.125549 446.296722 499.944611 533.982483 559.587830 580.415833 598.143738 613.687622 627.605286 ... 1156.336914 1167.988159 1180.645752 1194.563477 1210.107422 1227.835327 1248.663208 1274.268677 1308.306519 1361.954346
wa-vfr-noncity 2005-08-31 904.125549 440.744904 495.043365 529.493958 555.409851 576.490417 594.433289 610.165649 624.252136 ... 1159.395264 1171.187866 1183.999023 1198.085449 1213.817871 1231.760742 1252.841309 1278.757080 1313.207764 1367.506226

712 rows × 102 columns

Y_fitted_df
ds y AutoARIMA AutoARIMA-lo-98 AutoARIMA-lo-96 AutoARIMA-lo-94 AutoARIMA-lo-92 AutoARIMA-lo-90 AutoARIMA-lo-88 AutoARIMA-lo-86 ... AutoARIMA-hi-80 AutoARIMA-hi-82 AutoARIMA-hi-84 AutoARIMA-hi-86 AutoARIMA-hi-88 AutoARIMA-hi-90 AutoARIMA-hi-92 AutoARIMA-hi-94 AutoARIMA-hi-96 AutoARIMA-hi-98
unique_id
bus 1998-03-31 9815.0 9805.184570 7568.648926 7830.724121 7997.001953 8122.086426 8223.833008 8310.435547 8386.369141 ... 11037.260742 11094.178711 11156.011719 11224.000977 11299.934570 11386.537109 11488.283203 11613.368164 11779.646484 12041.720703
bus 1998-06-30 11823.0 11811.176758 9574.640625 9836.715820 10002.994141 10128.078125 10229.825195 10316.427734 10392.361328 ... 13043.252930 13100.169922 13162.003906 13229.993164 13305.926758 13392.528320 13494.275391 13619.360352 13785.637695 14047.712891
bus 1998-09-30 13565.0 13551.434570 11314.899414 11576.973633 11743.251953 11868.336914 11970.083008 12056.685547 12132.619141 ... 14783.510742 14840.428711 14902.261719 14970.250977 15046.184570 15132.787109 15234.533203 15359.618164 15525.896484 15787.970703
bus 1998-12-31 11478.0 11466.522461 9229.986328 9492.060547 9658.338867 9783.423828 9885.169922 9971.772461 10047.706055 ... 12698.597656 12755.515625 12817.348633 12885.337891 12961.271484 13047.874023 13149.620117 13274.705078 13440.983398 13703.057617
bus 1999-03-31 10027.0 9845.011719 7608.475586 7870.550781 8036.828613 8161.913086 8263.660156 8350.262695 8426.195312 ... 11077.086914 11134.004883 11195.838867 11263.828125 11339.760742 11426.363281 11528.110352 11653.194336 11819.472656 12081.547852
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
wa-vfr-noncity 2003-12-31 1177.0 927.351196 504.362732 553.928040 585.375671 609.032471 628.275513 644.654297 659.015320 ... 1160.369507 1171.134155 1182.828491 1195.687012 1210.048096 1226.426880 1245.669922 1269.326660 1300.774292 1350.339600
wa-vfr-noncity 2004-03-31 956.0 969.565552 546.577087 596.142456 627.590027 651.246887 670.489868 686.868652 701.229675 ... 1202.583862 1213.348511 1225.042847 1237.901489 1252.262451 1268.641235 1287.884277 1311.541016 1342.988647 1392.554077
wa-vfr-noncity 2004-06-30 772.0 967.268921 544.280457 593.845764 625.293396 648.950195 668.193237 684.572021 698.933044 ... 1200.287109 1211.051880 1222.746216 1235.604736 1249.965820 1266.344604 1285.587646 1309.244385 1340.692017 1390.257324
wa-vfr-noncity 2004-09-30 885.0 934.251831 511.263336 560.828674 592.276306 615.933105 635.176086 651.554932 665.915955 ... 1167.270020 1178.034790 1189.729126 1202.587646 1216.948730 1233.327515 1252.570557 1276.227295 1307.674927 1357.240234
wa-vfr-noncity 2004-12-31 797.0 925.923462 502.934998 552.500305 583.947937 607.604736 626.847778 643.226562 657.587585 ... 1158.941772 1169.706421 1181.400757 1194.259277 1208.620361 1224.999146 1244.242188 1267.898926 1299.346558 1348.911865

2492 rows × 103 columns

3. 调和预测结果

with CodeTimer('Reconcile Predictions ', verbose):
    if is_strictly_hierarchical(S=S_df.values.astype(np.float32), 
        tags={key: S_df.index.get_indexer(val) for key, val in tags.items()}):
        reconcilers = [
            BottomUp(),
            TopDown(method='average_proportions'),
            TopDown(method='proportion_averages'),
            MinTrace(method='ols'),
            MinTrace(method='wls_var'),
            MinTrace(method='mint_shrink'),
            #ERM(method='reg_bu', lambda_reg=100) # Extremely inneficient
            ERM(method='closed')
        ]
    else:
        reconcilers = [
            BottomUp(),
            MinTrace(method='ols'),
            MinTrace(method='wls_var'),
            MinTrace(method='mint_shrink'),
            #ERM(method='reg_bu', lambda_reg=100) # Extremely inneficient
            ERM(method='closed')
        ]
    
    hrec = HierarchicalReconciliation(reconcilers=reconcilers)
    Y_rec_df = hrec.bootstrap_reconcile(Y_hat_df=Y_hat_df,
                                        Y_df=Y_fitted_df,
                                        S_df=S_df, tags=tags,
                                        level=LEVEL,
                                        intervals_method=intervals_method,
                                        num_samples=10, num_seeds=10)

    # 匹配 Y_test/Y_rec/S 索引顺序
    Y_test_df = Y_test_df.reset_index()
    Y_test_df.unique_id = Y_test_df.unique_id.astype('category')
    Y_test_df.unique_id = Y_test_df.unique_id.cat.set_categories(S_df.index)
    Y_test_df = Y_test_df.sort_values(by=['unique_id', 'ds'])

    Y_rec_df = Y_rec_df.reset_index()
    Y_rec_df.unique_id = Y_rec_df.unique_id.astype('category')
    Y_rec_df.unique_id = Y_rec_df.unique_id.cat.set_categories(S_df.index)
    Y_rec_df = Y_rec_df.sort_values(by=['seed', 'unique_id', 'ds'])

    # 解析模型级列
    flat_cols = list(hrec.level_names.keys())
    for model in hrec.level_names:
        flat_cols += hrec.level_names[model]
    for model in hrec.sample_names:
        flat_cols += hrec.sample_names[model]
    y_rec  = Y_rec_df[flat_cols]
    model_columns = y_rec.columns

    n_series = len(S_df)
    n_seeds = len(Y_rec_df.seed.unique())
    y_rec  = y_rec.values.reshape(n_seeds, n_series, horizon, len(model_columns))
    y_test = Y_test_df['y'].values.reshape(n_series, horizon)
    y_train = Y_train_df['y'].values.reshape(n_series, -1)
Code block 'Reconcile Predictions ' took:   11.73492 seconds
# 解析分位数的定性评估
row_idx = 0
seed_idx = 0
col_idxs = model_columns.get_indexer(hrec.level_names['AutoARIMA/BottomUp'])
for i, col in enumerate(col_idxs):
    plt.plot(y_rec[seed_idx, row_idx,:,col], color='orange', alpha=i/100)
for i, col in enumerate(col_idxs):
    plt.plot(y_rec[seed_idx+1, row_idx,:,col], color='green', alpha=i/100)
plt.plot(y_test[row_idx,:], label='True')
plt.title(f'{S_df.index[row_idx]} Visits \n' + \
          f'AutoARIMA/BottomUp-{intervals_method}')

plt.legend()
plt.grid()
plt.show()
plt.close()

#Y_推荐数据框
td_levels = hrec.level_names['AutoARIMA/TopDown_method-average_proportions']
Y_rec_df[td_levels]
AutoARIMA/TopDown_method-average_proportions-lo-98 AutoARIMA/TopDown_method-average_proportions-lo-96 AutoARIMA/TopDown_method-average_proportions-lo-94 AutoARIMA/TopDown_method-average_proportions-lo-92 AutoARIMA/TopDown_method-average_proportions-lo-90 AutoARIMA/TopDown_method-average_proportions-lo-88 AutoARIMA/TopDown_method-average_proportions-lo-86 AutoARIMA/TopDown_method-average_proportions-lo-84 AutoARIMA/TopDown_method-average_proportions-lo-82 AutoARIMA/TopDown_method-average_proportions-lo-80 ... AutoARIMA/TopDown_method-average_proportions-hi-80 AutoARIMA/TopDown_method-average_proportions-hi-82 AutoARIMA/TopDown_method-average_proportions-hi-84 AutoARIMA/TopDown_method-average_proportions-hi-86 AutoARIMA/TopDown_method-average_proportions-hi-88 AutoARIMA/TopDown_method-average_proportions-hi-90 AutoARIMA/TopDown_method-average_proportions-hi-92 AutoARIMA/TopDown_method-average_proportions-hi-94 AutoARIMA/TopDown_method-average_proportions-hi-96 AutoARIMA/TopDown_method-average_proportions-hi-98
0 80750.389920 80750.389920 80750.389920 82299.061781 82299.061781 82299.061781 82600.022716 82600.022716 82600.022716 82763.007090 ... 88248.624229 88248.624229 88248.624229 88248.624229 88384.153447 90507.444522 90507.444522 90507.444522 90507.444522 90507.444522
1 61825.843210 61825.843210 61825.843210 61825.843210 61825.843210 61825.843210 63374.515072 63374.515072 63374.515072 63374.515072 ... 68196.499405 68196.499405 68286.705654 69324.077520 69324.077520 69437.018534 71582.897812 71582.897812 71582.897812 71582.897812
2 68249.624404 68249.624404 68249.624404 68249.624404 69798.296266 69798.296266 69798.296266 69798.296266 69798.296266 69798.296266 ... 74620.280598 74620.280598 74620.280598 74699.211067 75747.858714 75747.858714 75747.858714 75815.623322 78006.679006 78006.679006
3 67456.030661 67456.030661 67456.030661 67456.030661 69004.702523 69004.702523 69004.702523 69004.702523 69004.702523 69305.663457 ... 73939.444667 74954.264971 74954.264971 74954.264971 74954.264971 74954.264971 75044.617782 77213.085263 77213.085263 77213.085263
4 80371.819611 80371.819611 80371.819611 80371.819611 80371.819611 81827.571161 81920.491472 81920.491472 81920.491472 81920.491472 ... 87870.053920 87870.053920 87870.053920 87870.053920 88005.583138 90128.874213 90128.874213 90128.874213 90128.874213 90128.874213
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7115 132.959504 132.959504 132.959504 132.959504 132.959504 135.828870 136.012021 136.012021 136.012021 136.545910 ... 147.738932 147.738932 147.738932 152.191189 152.191189 152.191189 152.191189 152.191189 152.191189 152.191189
7116 158.417227 158.417227 158.417227 158.417227 158.417227 161.469743 161.469743 161.469743 161.469743 162.062953 ... 170.974136 171.174163 173.196654 173.196654 173.196654 173.419267 177.648912 177.648912 177.648912 177.648912
7117 122.066765 122.066765 125.119281 125.119281 125.119281 125.119281 125.712492 125.712492 125.712492 125.712492 ... 134.623675 134.623675 134.801476 136.846192 136.846192 136.846192 137.024283 141.298450 141.298450 141.298450
7118 134.467576 134.467576 134.467576 134.467576 137.520093 137.520093 137.520093 137.520093 138.113303 138.113303 ... 145.762241 145.875843 147.202288 149.247004 149.247004 149.247004 149.425094 153.699262 153.699262 153.699262
7119 135.996894 136.027420 136.027420 136.027420 136.027420 136.027420 136.027420 136.573173 136.620630 136.620630 ... 145.531813 145.531813 145.709614 147.754331 147.754331 147.754331 147.932421 152.206588 152.206588 152.206588

7120 rows × 100 columns

4. 评估

with CodeTimer('Evaluate Models CRPS  ', verbose):
    crps_results = {'Dataset': [dataset] * len(['Overall'] + list(tags.keys())),
                    'Level': ['Overall'] + list(tags.keys()),}

    for model in hrec.level_names.keys():
        crps_results[model] = []
        for level in crps_results['Level']:
            if level=='Overall':
                row_idxs = np.arange(len(S_df))
            else:
                row_idxs = S_df.index.get_indexer(tags[level])
            col_idxs = model_columns.get_indexer(hrec.level_names[model])
            _y = y_test[row_idxs,:]
            _y_rec_seeds = y_rec[:,row_idxs,:,:][:,:,:,col_idxs]

            level_model_crps = []
            for seed_idx in range(y_rec.shape[0]):
                _y_rec = _y_rec_seeds[seed_idx,:,:,:]
                level_model_crps.append(scaled_crps(y=_y, y_hat=_y_rec,
                                                    quantiles=QUANTILES))
            level_model_crps = f'{np.mean(level_model_crps):.4f}±{(1.96 * np.std(level_model_crps)):.4f}'
            crps_results[model].append(level_model_crps)

    crps_results = pd.DataFrame(crps_results)

crps_results
Code block 'Evaluate Models CRPS  ' took:   1.13514 seconds
Dataset Level AutoARIMA/BottomUp AutoARIMA/TopDown_method-average_proportions AutoARIMA/TopDown_method-proportion_averages AutoARIMA/MinTrace_method-ols AutoARIMA/MinTrace_method-wls_var AutoARIMA/MinTrace_method-mint_shrink AutoARIMA/ERM_method-closed_lambda_reg-0.01
0 TourismSmall Overall 0.0895±0.0012 0.1195±0.0008 0.1197±0.0008 0.0927±0.0010 0.0890±0.0010 0.0898±0.0009 0.1116±0.0015
1 TourismSmall Country 0.0481±0.0016 0.0479±0.0011 0.0479±0.0011 0.0504±0.0010 0.0510±0.0011 0.0512±0.0011 0.0525±0.0015
2 TourismSmall Country/Purpose 0.0699±0.0016 0.0928±0.0009 0.0931±0.0009 0.0804±0.0012 0.0724±0.0012 0.0741±0.0012 0.0927±0.0015
3 TourismSmall Country/Purpose/State 0.1085±0.0011 0.1575±0.0009 0.1579±0.0009 0.1082±0.0011 0.1043±0.0009 0.1049±0.0008 0.1325±0.0018
4 TourismSmall Country/Purpose/State/CityNonCity 0.1316±0.0012 0.1799±0.0008 0.1800±0.0008 0.1319±0.0013 0.1282±0.0011 0.1290±0.0010 0.1685±0.0029 
with CodeTimer('Evaluate Models MSSE  ', verbose):
    msse_results = {'Dataset': [dataset] * len(['Overall'] + list(tags.keys())),
                    'Level': ['Overall'] + list(tags.keys()),}
    for model in hrec.level_names.keys():
        msse_results[model] = []
        for level in msse_results['Level']:
            if level=='Overall':
                row_idxs = np.arange(len(S_df))
            else:
                row_idxs = S_df.index.get_indexer(tags[level])
            col_idx = model_columns.get_loc(model)
            _y = y_test[row_idxs,:]
            _y_train = y_train[row_idxs,:]
            _y_hat_seeds = y_rec[:,row_idxs,:,:][:,:,:,col_idx]

            level_model_msse = []
            for seed_idx in range(y_rec.shape[0]):
                _y_hat = _y_hat_seeds[seed_idx,:,:]
                level_model_msse.append(msse(y=_y, y_hat=_y_hat, y_train=_y_train))
            #level_model_msse = f'{np.mean(level_model_msse):.4f}±{(1.96 * np.std(level_model_msse)):.4f}'
            level_model_msse = f'{np.mean(level_model_msse):.4f}'
            msse_results[model].append(level_model_msse)

    msse_results = pd.DataFrame(msse_results)

msse_results
Code block 'Evaluate Models MSSE  ' took:   0.73303 seconds
Dataset Level AutoARIMA/BottomUp AutoARIMA/TopDown_method-average_proportions AutoARIMA/TopDown_method-proportion_averages AutoARIMA/MinTrace_method-ols AutoARIMA/MinTrace_method-wls_var AutoARIMA/MinTrace_method-mint_shrink AutoARIMA/ERM_method-closed_lambda_reg-0.01
0 TourismSmall Overall 0.2530 0.3628 0.3649 0.3039 0.2789 0.2822 0.3942
1 TourismSmall Country 0.2564 0.3180 0.3180 0.3522 0.3381 0.3394 0.4117
2 TourismSmall Country/Purpose 0.2018 0.3178 0.3203 0.2557 0.2122 0.2175 0.3346
3 TourismSmall Country/Purpose/State 0.3231 0.5077 0.5114 0.2943 0.2858 0.2890 0.4534
4 TourismSmall Country/Purpose/State/CityNonCity 0.3423 0.5047 0.5099 0.3238 0.3083 0.3115 0.4791 
with CodeTimer('Evaluate Models EScore', verbose):
    energy_results = {'Dataset': [dataset] * len(['Overall'] + list(tags.keys())),
                        'Level': ['Overall'] + list(tags.keys()),}
    for model in hrec.sample_names.keys():
        energy_results[model] = []
        for level in energy_results['Level']:
            if level=='Overall':
                row_idxs = np.arange(len(S_df))
            else:
                row_idxs = S_df.index.get_indexer(tags[level])
            col_idxs = model_columns.get_indexer(hrec.sample_names[model])
            _y = y_test[row_idxs,:]
            _y_sample1 = y_rec[0,row_idxs,:,:][:,:,col_idxs[:len(col_idxs)//2]]
            _y_sample2 = y_rec[0,row_idxs,:,:][:,:,col_idxs[len(col_idxs)//2:]]
            level_model_energy = energy_score(y=_y, 
                                              y_sample1=_y_sample1,
                                              y_sample2=_y_sample2,
                                              beta=2)
            energy_results[model].append(level_model_energy)
    energy_results = pd.DataFrame(energy_results)

energy_results
Code block 'Evaluate Models EScore' took:   0.19443 seconds
Dataset Level AutoARIMA/BottomUp AutoARIMA/TopDown_method-average_proportions AutoARIMA/TopDown_method-proportion_averages AutoARIMA/MinTrace_method-ols AutoARIMA/MinTrace_method-wls_var AutoARIMA/MinTrace_method-mint_shrink AutoARIMA/ERM_method-closed_lambda_reg-0.01
0 TourismSmall Overall 6.874103e+07 7.917294e+07 7.962361e+07 6.930268e+07 6.914837e+07 6.955018e+07 8.235776e+07
1 TourismSmall Country 3.292999e+07 2.757131e+07 2.757129e+07 3.081254e+07 3.392861e+07 3.353851e+07 3.350023e+07
2 TourismSmall Country/Purpose 1.894485e+07 2.661024e+07 2.683828e+07 2.218952e+07 1.932895e+07 1.984161e+07 2.681792e+07
3 TourismSmall Country/Purpose/State 9.393103e+06 1.408613e+07 1.419471e+07 9.016056e+06 8.778983e+06 8.928542e+06 1.211747e+07
4 TourismSmall Country/Purpose/State/CityNonCity 7.473085e+06 1.090527e+07 1.101934e+07 7.284562e+06 7.111832e+06 7.241519e+06 9.922145e+06

参考文献

If you find the code useful, please ⭐ us on Github