Fork me on GitHub
#ai
<
2022-09-16
>
kwladyka13:09:15

https://unit8co.github.io/darts/generated_api/darts.models.forecasting.rnn_model.html Does backtest() do what I think? I have hard time to understand this. I need to use test dataset to: 1. predict 1 value ahead 2. train model with this +1 value using test dataset 3. loop On the end I want to plot chart with both real values from test dataset and predicted values to compare. The whole point is I want to simulate how model behave predicting only 1 value ahead and training with real value from dataset. Not the value which were predicted. In other words the purpose of the model is to predict only 1 value ahead, not more. Because the model will know real value in each minute and I want to predict next minute. Ah I hope I made it clear :)

kwladyka13:09:17

if not what I can use or I have to code it myself?

kwladyka13:09:13

I think backtest doesn’t do what I want

kwladyka13:09:09

backtest vs histroical_forecast vs residuals Is it clear for you which one I should use?

kwladyka19:09:24

after all I think histroical_forecast