RNN - Recurrent Neural Network

An RNN used for time series forecasting is a Machine Learning model, and more specifically, it is a type of Deep Learning model. RNNs are a class of neural networks well-suited for time series forecasting, especially when dealing with sequential data where previous values influence future ones.

🔁How RNN Works for Time Series Forecasting

• Sequential Processing: Unlike feedforward neural nets, RNNs have loops that allow information to persist, making them ideal for temporal patterns.

• Memory of Past Inputs: RNNs maintain a hidden state that gets updated as new inputs arrive—ideal for learning trends and seasonality in time series.

• Prediction Strategy:

  • One-to-One: Predict the next value.

  • Many-to-One: Use a sequence of values to predict one output.

  • Many-to-Many: Sequence in, sequence out (like stock price forecast over the next week).

RNN Model Parameters

Param Name	Description	Param Value (Default)
RNN-TYPE	Type of RNN architecture used (e.g., LSTM, GRU, RNN).	LSTM
HIDDEN-DIM	Number of hidden units in each RNN layer.	124
RNN-LAYERS	Number of stacked RNN layers in the model.	2
DROPOUT	Dropout rate applied between RNN layers to prevent overfitting.	0.0
EPOCHS	Total number of times the model trains on the dataset.	100
LEARNING-RATE	Learning rate used by the optimizer for training.	0.001
BATCH-SIZE	Number of training samples per batch.	32
RANDOM-STATE	Seed value for reproducibility of model training.	42
FORCE-RESET	Whether to reset hidden states between sequences (for stateful RNNs).	True