Arrival Time Prediction for Public Transport Using LSTM-Based Neural Networks

Neural networks have recently found widespread application across various domains within IT software infrastructure. This study focuses on specific neural network architectures incorporating Long Short-Term Memory (LSTM) layers and their variations, aiming to effectively capture the dynamic, nonlinear nature of traffic data. LSTM networks are well-suited for learning long-term dependencies in sequential data, making them particularly effective for time series prediction tasks. To evaluate the performance of different LSTM-based architectures, we utilize a dataset comprising bus logs from New York City collected over a four-month period. The experimental results indicate that LSTM architectures demonstrate strong predictive capabilities and are well-suited for modeling complex temporal patterns in traffic data.