I have a dataset of 1000 rows (observations) and 100 columns (features). My features are all time-dependent features at different equal time steps, and my output is the last timestep (last column).
I want to use the 1D-CNN for the dataset. However, I have a problem once any model is built on this dataset, I only can provide the initial step of my feature (first column) for the prediction. in another word, I want to give the initial step and predict the next steps through time.
I have seen to use the 1D-CNN in normal problems we need to give a sequence of inputs (as we used in training) to predict the next step.
I wondered to know if I can use the 1D-CNN for such a case or I have to look for other models such as LSTM or others?
question from:
https://stackoverflow.com/questions/65942284/can-we-use-1d-cnn-for-time-series-once-only-the-first-step-is-available-for-pred 与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…