I have a sequence of images, let's say we ignore time specificity for now. In the other hand, target is a multivariate continuous time series. Let's consider it just a univariate one. Training a cnn model didn't got me a satisfying results. A very sure reason is that train is only on one year, but also, there are some missing. While forecast is one several months knowing their images and ignoring their target. I thought I could add time features like trends and seasonality. I thought just adding very useful ones is better since the length of dataset is not that huge relatively with horizon. Doing that seems challenging. I read about GAN models, and thought features could be generated from train images and target paralelly. This model will predict from images those features. Another one will use both as inputs to finally predict target .

My question now is: is this doable?. Is it really worth trying? And any hints on implemented models already. Please let me know if anything is not that clear. And bear with my formulation of this problem as I'm a noob in deep learning and not supposed to be an expert.


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.