Issue Regarding Multiple Step Forecasting #5
Description
Hi Petrônio,
I am currently working on a real-world problem which is to forecast the next 24-hrs demand. However, when I try to run model.predict(input_data, steps_ahead=24), it only gives me constant values that are not reasonable.
I referred to the article you published and the graph shows the results look quite nice. I am keen to know how to achieve this multi-step forecasting. Are there any pre-processing steps needed for the input test data before predicting？
In addition, I noticed that some of the examples you gave as following:
ax[count].plot(dataset[train_split:train_split+200])
model1 = cUtil.load_obj('model1'+dataset_name+str(order))
forecasts = model1.predict(dataset[train_split:train_split+200])
ax[count].plot(forecasts)
From my understanding, if the input test data is from time t, the forecasts should be starting from time (t+1). We need to use the previous data to predict the next one. Then ax[count].plot(dataset[train_split:train_split+200]) should be changed to ax[count].plot(dataset[train_split+1:train_split+201]), right?
Looking forward to your reply.!
Thank you so much!