Evaluating The Performance Of Lstm In Traffic Flow Prediction At Different Time Scales
Keywords:
Recurrent Neural Networks (RNN), Gated Recurrent Units (GRU), and Long Short-Term Memory (LSTM)Abstract
This project presents a traffic prediction system leveraging deep learning techniques such as Recurrent Neural Networks (RNN), Gated Recurrent Units (GRU), and Long Short-Term Memory (LSTM) networks to analyze and forecast vehicular traffic volume based on historical time-series data. Developed using Python and integrated within a Django web framework, the system utilizes traffic data that includes timestamp information to train models and predict traffic patterns across different hours of the day. The dataset undergoes preprocessing steps including feature extraction from timestamps, normalization using MinMaxScaler, and sequence generation for time-series forecasting. The models are trained using the Keras deep learning library, and their performance is evaluated using the Root Mean Squared Error (RMSE) metric. Among the models, LSTM has been configured with multiple hidden layers and dropout regularization to handle temporal dependencies and avoid overfitting. Users interact with the system through a web interface that allows them to initiate training for each model (RNN, GRU, LSTM) and visualize prediction outputs in comparison to actual traffic data. The system also includes a prediction module where users can simulate hourly traffic forecasts for a given year, with traffic levels categorized into “Low”, “Mild”, or “High” for intuitive understanding. This intelligent traffic forecasting tool can aid in smarter city planning, traffic control, and congestion management by offering accurate, real-time, and data-driven predictions.
Downloads
References
1. Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735–1780.This foundational paper introduces the LSTM architecture, a type of recurrent neural network that can learn long-term dependencies, which is extensively used in traffic prediction tasks.
2. Cho, K., van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation. arXiv preprint arXiv:1406.1078.This paper presents the GRU (Gated Recurrent Unit), an efficient and effective variant of LSTM that requires fewer resources, ideal for time-series prediction like traffic forecasting.
3. Zhang, Y., Liu, Y., & Wang, L. (2019). Road Traffic Flow Prediction Model Based on Long Short-Term Memory Neural Network. Procedia Computer Science, 154, 377–384.Discusses the application of LSTM in traffic flow prediction using historical datasets, showcasing its advantages in capturing temporal traffic patterns.
4. Ma, X., Tao, Z., Wang, Y., Yu, H., & Wang, Y. (2015). Long short-term memory neural network for traffic speed prediction using remote microwave sensor data. Transportation Research Part C: Emerging Technologies, 54, 187–197.This research evaluates the effectiveness of LSTM in predicting traffic speed, utilizing real-time sensor data, closely aligning with the goals of this system.
5. Yu, B., Song, H., & Guan, H. (2016). A deep learning approach to urban traffic flow prediction based on time-series data. Big Data Analytics, 1(1), 1–10Explores deep learning techniques including RNN and LSTM for urban traffic forecasting, supporting the comparative methodology used in this project.
