In a time-series prediction task, how does an RNN effectively model sequential data?

RNNs model sequential data in time-series tasks by maintaining hidden states that propagate information across time steps, capturing temporal dependencies like trends or seasonality. This memory mechanism allows RNNs to learn from past data, unlike independent processing or holistic approaches, though they face gradient issues for long sequences. Exact extract: "RNNs use hidden states to retain context from prior time steps, effectively capturing dependencies in sequential data for time-series tasks." (Reference: Cyber Security for AI by SISA Study Guide, Section on RNN Architectures, Page 40-43).