Hello to evereone I have a li ttle trouble with prediction i

Question

Hello to evereone I have a li ttle trouble with prediction intervals here The attached file contains my dataframe with 924 unique ids and some static variables Follow my code ```train = df completo encoded query ds lt 2023 07 07 test = df completo encoded query ds gt = 2023 07 07 from mlforecast utils import PredictionIntervals from xgboost import XGBRegressor from lightgbm import LGBMRegressor from mlforecast target transforms import Differences from mlforecast utils import PredictionIntervals from window ops ewm import ewm mean from window ops rolling import rolling mean seasonal rolling mean rolling min rolling max rolling std mlf = MLForecast freq = D models= XGBRegressor n jobs = 1 LGBMRegressor n jobs = 1 target transforms= Differences 1 7 lag transforms= 1 rolling mean 2 rolling mean 3 rolling mean 4 rolling mean 5 rolling mean 6 rolling mean 7 rolling mean 7 rolling mean 14 rolling mean 28 ewm mean 0 9 expanding mean rolling min 7 rolling min 14 rolling min 28 rolling max 7 rolling max 14 rolling max 28 rolling std 2 rolling std 3 rolling std 4 rolling std 5 rolling std 6 rolling std 7 rolling std 14 rolling std 28 lags= 1 7 14 21 28 date features= month year day of week day of year is month start quarter days in month num threads=4 time mlf fit train id col= unique id max horizon = 47 prediction intervals=PredictionIntervals n windows=10 window size=47 time col= ds target col= y static features= gtin ADI CV2 cluster 0 cluster 1 cluster 2 cluster 3 cluster 4 ``` To here everything ok But follow de error ```levels = 50 80 95 forecasts = mlf predict 47 level = levels NotImplementedError Traceback most recent call last Cell In 81 line 3 1 levels = 50 80 95 gt 3 forecasts = mlf predict 47 level = levels 4 forecasts head File usr local lib python3 10 site packages mlforecast utils py 186 in old kw to pos lt locals gt decorator lt locals gt inner args kwargs 184 new args append kwargs pop arg names i 185 new args append kwargs pop old name gt 186 return f new args kwargs File usr local lib python3 10 site packages mlforecast forecast py 532 in MLForecast predict self h dynamic dfs before predict callback after predict callback new df level X df ids horizon new data 528 model names = self models keys 529 conformal method = get conformal method 530 self prediction intervals method 531 gt 532 forecasts = conformal method 533 forecasts 534 self cs df 535 model names=list model names 536 level=level 537 cs h=self prediction intervals h 538 cs n windows=self prediction intervals n windows 539 n series=self ts ga ngroups 540 horizon=h 541 542 return forecasts File usr local lib python3 10 site packages mlforecast forecast py 55 in add conformal distribution intervals fcst df cs df model names level cs n windows cs h n series horizon 53 scores = scores horizon 54 mean = fcst df model values reshape 1 n series 1 gt 55 scores = np vstack mean scores mean + scores 56 quantiles = np quantile 57 scores 58 cuts 59 axis=0 60 61 quantiles = quantiles reshape len cuts 1 File usr local lib python3 10 site packages pandas core arrays masked py 528 in BaseMaskedArray array ufunc self ufunc method inputs kwargs 525 return NotImplemented 527 for binary ops use our custom dunder methods gt 528 result = ops maybe dispatch ufunc to dunder op 529 self ufunc method inputs kwargs 530 531 if result is not NotImplemented 532 return result File usr local lib python3 10 site packages pandas libs ops dispatch pyx 113 in pandas libs ops dispatch maybe dispatch ufunc to dunder op File usr local lib python3 10 site packages pandas core ops common py 81 in unpack zerodim and defer lt locals gt new method self other 77 return NotImplemented 79 other = item from zerodim other gt 81 return method self other File usr local lib python3 10 site packages pandas core arraylike py 198 in OpsMixin rsub self other 196 unpack zerodim and defer rsub 197 def rsub self other gt 198 return self arith method other roperator rsub File usr local lib python3 10 site packages pandas core arrays masked py 659 in BaseMaskedArray arith method self other op 657 other = np asarray other 658 if other ndim gt 1 gt 659 raise NotImplementedError can only perform ops with 1 d structures 661 We wrap the non masked arithmetic logic used for numpy dtypes 662 in Series Index arithmetic ops 663 other = ops maybe prepare scalar for op other len self NotImplementedError can only perform ops with 1 d structures```

José Morales · Answer

Hey Did you run the fit first with max horizon

Mairon Cesar Simoes Chaves · Answer

Hello José The max horizon in the fit argument is commented out I didn t use it Should max horizon influence

José Morales · Answer

I think if you run it first with and then without you can run into that error if you don t re create the MLForecast object

Mairon Cesar Simoes Chaves · Answer

i restarted the kernel and tried again The training even goes ok the message occurs is at the time of the predict ```mlf fit train id col= unique id prediction intervals=PredictionIntervals n windows=10 window size=47 time col= ds target col= y static features= gtin ADI CV2 cluster 0 cluster 1 cluster 2 cluster 3 cluster 4 LightGBM Warning Auto choosing row wise multi threading the overhead of testing was 0 063419 seconds You can set `force row wise=true` to remove the overhead And if memory is not enough you can set `force col wise=true` LightGBM Info Total Bins 5559 LightGBM Info Number of data points in the train set 378531 number of used features 44 LightGBM Info Start training from score 0 000629 LightGBM Warning Auto choosing row wise multi threading the overhead of testing was 0 088819 seconds You can set `force row wise=true` to remove the overhead And if memory is not enough you can set `force col wise=true` LightGBM Info Total Bins 5930 LightGBM Info Number of data points in the train set 811401 number of used features 44 LightGBM Info Start training from score 0 000392 CPU times user 13min 1s sys 4 26 s total 13min 5s Wall time 3min 55s 37 MLForecast models= XGBRegressor LGBMRegressor freq= lt Day gt lag features= lag1 lag7 lag14 lag21 lag28 rolling mean lag1 window size2 rolling mean lag1 window size3 rolling mean lag1 window size4 rolling mean lag1 window size5 rolling mean lag1 window size6 rolling mean lag1 window size7 rolling mean lag1 window size14 rolling mean lag1 window size28 ewm mean lag1 alpha0 9 expanding mean lag1 rolling min lag1 window size7 rolling min lag1 window size14 rolling min lag1 window size28 rolling max lag1 window size7 rolling max lag1 window size14 rolling max lag1 window size28 rolling std lag1 window size2 rolling std lag1 window size3 rolling std lag1 window size4 rolling std lag1 window size5 rolling std lag1 window size6 rolling std lag1 window size7 rolling std lag1 window size14 rolling std lag1 window size28 date features= month year day of week day of year is month start quarter days in month num threads=4 ``` ```mlf prediction intervals PredictionIntervals n windows=10 h=47 method= conformal distribution ``` ```levels = 50 80 95 forecasts = mlf predict 47 level = levels NotImplementedError Traceback most recent call last Cell In 42 line 3 1 levels = 50 80 95 gt 3 forecasts = mlf predict 47 level = levels 4 forecasts head File usr local lib python3 10 site packages mlforecast utils py 186 in old kw to pos lt locals gt decorator lt locals gt inner args kwargs 184 new args append kwargs pop arg names i 185 new args append kwargs pop old name gt 186 return f new args kwargs File usr local lib python3 10 site packages mlforecast forecast py 532 in MLForecast predict self h dynamic dfs before predict callback after predict callback new df level X df ids horizon new data 528 model names = self models keys 529 conformal method = get conformal method 530 self prediction intervals method 531 gt 532 forecasts = conformal method 533 forecasts 534 self cs df 535 model names=list model names 536 level=level 537 cs h=self prediction intervals h 538 cs n windows=self prediction intervals n windows 539 n series=self ts ga ngroups 540 horizon=h 541 542 return forecasts File usr local lib python3 10 site packages mlforecast forecast py 55 in add conformal distribution intervals fcst df cs df model names level cs n windows cs h n series horizon 53 scores = scores horizon 54 mean = fcst df model values reshape 1 n series 1 gt 55 scores = np vstack mean scores mean + scores 56 quantiles = np quantile 57 scores 58 cuts 59 axis=0 60 61 quantiles = quantiles reshape len cuts 1 File usr local lib python3 10 site packages pandas core arrays masked py 528 in BaseMaskedArray array ufunc self ufunc method inputs kwargs 525 return NotImplemented 527 for binary ops use our custom dunder methods gt 528 result = ops maybe dispatch ufunc to dunder op 529 self ufunc method inputs kwargs 530 531 if result is not NotImplemented 532 return result File usr local lib python3 10 site packages pandas libs ops dispatch pyx 113 in pandas libs ops dispatch maybe dispatch ufunc to dunder op File usr local lib python3 10 site packages pandas core ops common py 81 in unpack zerodim and defer lt locals gt new method self other 77 return NotImplemented 79 other = item from zerodim other gt 81 return method self other File usr local lib python3 10 site packages pandas core arraylike py 198 in OpsMixin rsub self other 196 unpack zerodim and defer rsub 197 def rsub self other gt 198 return self arith method other roperator rsub File usr local lib python3 10 site packages pandas core arrays masked py 659 in BaseMaskedArray arith method self other op 657 other = np asarray other 658 if other ndim gt 1 gt 659 raise NotImplementedError can only perform ops with 1 d structures 661 We wrap the non masked arithmetic logic used for numpy dtypes 662 in Series Index arithmetic ops 663 other = ops maybe prepare scalar for op other len self NotImplementedError can only perform ops with 1 d structures```

José Morales · Answer

Are you on the latest version This runs fine for me ```from mlforecast import MLForecast from mlforecast utils import PredictionIntervals from xgboost import XGBRegressor from lightgbm import LGBMRegressor from mlforecast target transforms import Differences from mlforecast utils import PredictionIntervals generate daily series from window ops ewm import ewm mean from window ops rolling import rolling mean seasonal rolling mean rolling min rolling max rolling std from window ops expanding import expanding mean train = generate daily series 10 min length=1 000 max length=1 000 n static features=2 static as categorical=False mlf = MLForecast freq = D models= XGBRegressor n jobs = 1 LGBMRegressor n jobs = 1 target transforms= Differences 1 7 lag transforms= 1 rolling mean 2 rolling mean 3 rolling mean 4 rolling mean 5 rolling mean 6 rolling mean 7 rolling mean 7 rolling mean 14 rolling mean 28 ewm mean 0 9 expanding mean rolling min 7 rolling min 14 rolling min 28 rolling max 7 rolling max 14 rolling max 28 rolling std 2 rolling std 3 rolling std 4 rolling std 5 rolling std 6 rolling std 7 rolling std 14 rolling std 28 lags= 1 7 14 21 28 date features= month year day of week day of year is month start quarter days in month num threads=4 mlf fit train id col= unique id max horizon = 47 prediction intervals=PredictionIntervals n windows=10 window size=47 time col= ds target col= y static features= static 0 static 1 mlf predict 47 level = 50 80 95 ```

Mairon Cesar Simoes Chaves · Answer

Hello Jose Thanks for the reply I m investigating if it could be something in the sagemaker environment I m running I ll be back soon with more information

José Morales · Answer

Also if you re able to print the shapes of the means and scores it d help a lot If you run ` debug` on your notebook when you get the error you can go up with `u` and get to that point and then just print the shapes I think it could be because some small series as well

Mairon Cesar Simoes Chaves · Answer

Yes I m investigating whether it could be the size of the series or also whether during cross validation it s getting a horizon in which all observations are zeros given the intermittence of the skus here at the company

Mairon Cesar Simoes Chaves · Answer

José maybe can you help me with a thing here I m currently facing a challenge in my e commerce business which comprises a diverse range of SKUs each with its own unique sales behavior and patterns Our goal is to make demand forecasts for the upcoming 30 days however due to the diversity and complexity of the data finding an efficient strategy to select relevant and predictable SKUs has proven to be quite a challenge To date we ve gathered sales data from January 2021 to the present date in August 2023 encompassing over a million GTINs Global Trade Item Numbers with sales records This massive number of time series data points prompts us to seek an approach that s both effective and manageable One of the approaches we re considering is applying the Pareto principle to prioritize the GTINs By doing so we re able to reduce the number of SKUs from over a million to around 20 000 This initial selection allows us to concentrate our efforts on the most promising time series data Among the behavior patterns we ve identified are 1 Single Sales and Halting Some GTINs sell only a single unit and then halt sales completely 2 Sales Decline Other GTINs had a strong sales history in the past but have stopped entirely for months 3 Extreme Intermittency Certain GTINs exhibit extremely intermittent patterns selling one unit one day going days without sales and occasionally selling one unit again Many of these cases show minimal variance in the data with the time series inflated by zeros on days without sales 4 New Launches Other GTINs have recently begun selling and have limited sales history 5 Variable Sales There are GTINs that sell one unit on alternate days but at specific times suddenly register extremely high sales followed by a sales interruption 6 Diverse Time Periods Additionally the time series of GTINs have varying start and end dates adding an extra layer of complexity Considering these behavior patterns I m seeking guidance on the best approach to selecting the most relevant and predictable GTINS for our demand forecasting efforts Have you experienced a similar situation where you had to filter out predictable products Your insights advice and shared experiences would be immensely valuable in refining our strategy and optimizing our outcomes

Mairon Cesar Simoes Chaves · Answer

in some experiments I even get very interesting predictions according to the graph But we still need to add the confidence interval and deep down I still doubt whether the strategy to select the gtins was the best

José Morales · Answer

hey Another possible approach would be using with some dummy models like Naive and SeasonalNaive and a more sophisticated one like ARIMA or ETS and see if they outperform the dummy ones Where they don t you could say that there isn t enough signal to forecast them Also since the volume you have is high you could try as well shows how you can compare them and even combine them

Mairon Cesar Simoes Chaves · Answer

cool and for the neuralforecast do you recommend balancing the panel leaving all the time series with the same length or is it not necessary

José Morales · Answer

Only if they re too short for your windows We have the <https nixtla github io utilsforecast preprocessing html fill gaps|fill gaps> function which may be of help