why use previous observed data as input instead of previous prediction?

sunfanyunn / DARNN

A Dual-Stage Attention-Based Recurrent Neural Network for Time Series Prediction

112 stars 23 forks source link

why use previous observed data as input instead of previous prediction? #4

Open AtrCheema opened 4 years ago

AtrCheema commented 4 years ago

May be a very basic question as I may be missing something. Why use the observed data at previous step to predict next value? What kind of prediction model is this when we still need observations to make predictions during "prediction"/deployment stage?