Simple RNNs for time series data

In this recipe, we will learn how to use a simple RNN implementation of Keras to predict sales based on a historical dataset.

Note

RNNs are a class of artificial neural network where connections between nodes of the network form a directed graph along a sequence. This topology allows it to exhibit dynamic temporal behavior for input of the time sequence type. Unlike feedforward neural networks, RNNs can use their internal state (also calledmemory) to process sequences of inputs. This makes them suitable for tasks such as unsegmented, connected handwriting recognition or speech recognition.

A simple RNN is implemented as part of the keras.layers.SimpleRNN class as follows:

keras.layers.SimpleRNN(units, activation='tanh', 
   use_bias=True, 
   kernel_initializer='glorot_uniform', 
   recurrent_initializer='orthogonal', 
   bias_initializer='zeros', 
   kernel_regularizer=None, 
   recurrent_regularizer=None, 
   bias_regularizer=None, 
   activity_regularizer...