How to get a better forecast with DeepAr?

[Msaleh87]

I have an electric load data for about a year, I trained the model for a data set from Mar 2021 until Mar 2022. I only used one feature (zeros and ones for weekends and weekdays). Here is my code:

estimator = DeepAREstimator(freq = '5min',
                           context_length = 7*24*12,  #The previous data points I look at to predict. It's like n_forecast in Neural Prophet 
                           prediction_length = 1*24*12,
                           num_layers = 2,
                           num_cells = 64,
                           cell_type = 'lstm',
                           use_feat_dynamic_real =True,
                           trainer = Trainer(epochs= 50))

from gluonts.dataset.common import ListDataset
training_data = ListDataset([{"start": df_input.index[0],
                              "target": df_input.Load[:'2022-03-07 04:50:00'],
                              'feat_dynamic_real': [df_input.DofW[:'2022-03-07 04:50:00']]}],
                           freq = "5min")

predictor = estimator.train(training_data = training_data)

test_data = ListDataset(
    [
        {"start": df_input.index[0], "target": df_input.Load[:'2022-03-08 04:50:00'], 'feat_dynamic_real': [df_input.DofW[:'2022-03-08 04:50:00']]}, # one day
        {"start": df_input.index[0], "target": df_input.Load[:'2022-03-09 04:50:00'], 'feat_dynamic_real': [df_input.DofW[:'2022-03-09 04:50:00']]}   # two days
    
    ],

freq = "5min")

forecast_it, ts_it = make_evaluation_predictions(
    dataset = test_data,
    predictor = predictor,
    num_samples = 1000 # I used 100 as well and it didn't make a big difference
)

These are the results I got, after almost two hours of training. What hyperparameters I should add/ tune to make the forecast better? I am hoping to use it to predict 48 hours ahead. Also, I used Neural Prophet and it gave me way better results for 48 hours ahead so far.

[Msaleh87]

That question was answered here #2242 by @lostella