How to extract and plot the forecasted data from MQ-CNN model

[Msaleh87]

So I was trying to plot the forecast data on top of the test data. I can plot them separately, but not on top of each other. Then I tried to extract the forecast data, but for some reason, I could not. Here is my code:

from gluonts.model.seq2seq import MQCNNEstimator
estimator = MQCNNEstimator(freq = '5min',
                           #context_length = 2*24*12,  
                           prediction_length = 1*24*12,
                           scaling = True,
                           use_feat_dynamic_real = True,
                           trainer = Trainer(epochs= 4, batch_size = 32 * 20 ,num_batches_per_epoch= int(3141/20) ,learning_rate = 0.001,hybridize=True, callbacks=[scheduler]))


test_data = ListDataset(
    [
        {"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']], 
        'time_feat': [time_feat_df.DayofWeek[:'2022-03-09 04:50:00'],
                     time_feat_df.Hour[:'2022-03-09 04:50:00'],]}, 

        {"start": df_input.index[0], "target": df_input.Load[:'2022-03-10 04:50:00'], 
         'feat_dynamic_real': [df_input.DofW[:'2022-03-10 04:50:00']],
        'time_feat': [time_feat_df.DayofWeek[:'2022-03-10 04:50:00'],
                    time_feat_df.Hour[:'2022-03-10 04:50:00'],]}   
    
    ],

freq = "5min")

from gluonts.evaluation.backtest import make_evaluation_predictions
forecast_it, ts_it = make_evaluation_predictions(
    dataset = test_data,
    predictor = predictor,
    num_samples = 20
)

forecasts = list(forecast_it)
tss = list(ts_it)

Now if I want to plot I do:

forecasts[0].plot()
tss[0][-288:].plot() 

How can plot them on top of each other, and more importantly how can I extract the forecast data in the format of timestamps and corresponding values? Every time I try any command on forecast[0], it gives me an error that it's a QuantileForecast object
QuantileForecast' object has no attribute

I checked this one, but didn't help: #987

[Msaleh87]

Hi @mbohlkeschneider
We discussed MQ-CNN before. I was wondering if you could help regarding this question. Whenever I write the command forecast[0], I get something like forecast[0] = QuantileForecast(array[... forecast numbers ...] , ..., quantiles, forecasts_keys), and I am not sure how to extract the forecast data.

[Msaleh87]

So here is what I came up with to extract and plot the data from MQ-CNN model and I hope this help:

First, step up the test data:

#Specify the end dates you like for the forecast
test_data = ListDataset(
    [
        {"start": df_input.index[0], "target": df_input.Load[:'end_date'], 
         'feat_dynamic_real': [df_input.DofW[:'end_date']], 
        'time_feat': [time_feat_df.DayofWeek[:'end_date'],
                     time_feat_df.Hour[:'end_date'],]},
        {"start": df_input.index[0], "target": df_input.Load[:'end_date2'], 
         'feat_dynamic_real': [df_input.DofW[:'end_date2']],
        'time_feat': [time_feat_df.DayofWeek[:'end_date2'],
                    time_feat_df.Hour[:'end_date2'],]}  
    
    ],

freq = "5min")

from gluonts.evaluation.backtest import make_evaluation_predictions
forecast_it, ts_it = make_evaluation_predictions(
    dataset = test_data,
    predictor = predictor,
    num_samples = 20 # up to you to decide 10-->100 usually works
)

forecasts = list(forecast_it) # It has 2 sets of forecasts since you had 2 sets of test datasets
tss = list(ts_it)

Second, extracting and plotting the data:

ts_e = tss[0]
forecast_e = forecasts[0]
f_length = len(forecast_e.mean)

data = {
        'timestamp': ts_e.index[-f_length:].strftime("%Y-%m-%d %H:%M:%S"),
        'forecast': forecast_e.mean, # this is the forecasted data # <----------
        'Real data' : ts_e[-f_length:][0].values
    }
dff = pd.DataFrame(data)

dff = dff.set_index('timestamp')
dff.plot() # it will plot the real and the forecasted data