Building an RNN network in Keras is much simpler as compared to building using lower=level TensorFlow classes and methods. For Keras, we preprocess the data, as described in the previous sections, to get the supervised machine learning time series datasets: X_train, Y_train, X_test, Y_test.
From here onwards, the preprocessing differs. For Keras, the input has to be in the shape (samples, time steps, features). As we converted our data to the supervised machine learning format, while reshaping the data, we can either set the time steps to 1, thus feeding all input time steps as features, or we can set the time steps to the actual number of time steps, thus feeding the feature set for each time step. In other words, the X_train and X_test datasets that we obtained earlier could be reshaped as one of the following methods:
Method...
