Let's say we have 8,000 different time series where each of them has 10,000 samples and 25 features. The goal is to have an LSTM sequence to sequence model (using Keras) where one can use a sequence of 7 elements for each of these time series in order to predict the following 4 elements of that series. I believe to do this for one of the time series: one has to create the 7 lags of the series and the 7 lags of each of the features and the input data would have the shape (10000, 7, 25x7). How would the input and output shapes look like if one wanted to create a single model for all the series at the same time?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.