LSTM for Traffic Speed Prediction

Memory Beneath the Speed Curve

Traffic speed has a memory. A slowdown observed now may be part of a brief fluctuation or the beginning of a longer queue, and those possibilities look different when recent history is taken into account. Short-term prediction is therefore a question of which past observations should remain influential.

An evolving speed series supplies the network with observations one moment at a time. Its internal memory changes as each new reading arrives: some information is retained, some is replaced, and the resulting state is used to forecast the next speed. Comparing predicted and observed values shows both the promise and the limits of learning from a sequence. The core insight is selective memory: the model is designed to carry useful context across time rather than treating every reading as an unrelated snapshot. That makes it a natural tool for traffic states whose changes unfold over many measurement intervals.