Short-Term Traffic Speed Forecasting Using a Deep Learning Method Based on Multitemporal Traffic Flow Volume

Abstract

Accurate traffic speed forecasting not only can help traffic management departments make better judgments and improve the efficacy of road monitoring but also can help drivers plan their driving routes and arrive safely and smoothly at their destination. This paper focuses on the lack of traffic speed data and proposes a method for traffic speed forecasting based on the multitemporal traffic flow volume of the previous and later moment states. First, according to traffic flow volume data, the different traffic patterns of previous and later moment states were extracted. Second, the performance of five forecasting models, namely, long short-term memory (LSTM), backpropagation (BP), classification and regression trees, k-nearest neighbor, and support vector regression, were compared. Finally, the model with the best prediction results was used to conduct sensitivity analysis experiments for different traffic patterns. Through a real-data case study, we found that the LSTM model has the highest prediction accuracy compared to other models in both time and space. This traffic pattern "previous = 3 and later = 3" can forecast traffic speed more accurately, and its forecasting ability is robust across a range of scenarios.

Enhanced Deep Neural Networks for Traffic Speed Forecasting Regarding Sustainable Traffic Management Using Probe Data from Registered Transport Vehicles on Multilane Roads

“…Normally, PSO-LSTM and GA-LSTM optimized the following effect parameters to make the model more efficient, such as the window size, the number of epochs, the number of neurons, and the learning rate. In the introduction section, in the previously published research [7], we only optimized three effective parameters: the window size, the number of epochs, and the number of neurons, respectively, similar to applying the LSMT network to optimize forecasting models in several studies [27][28][29]. This causes a disadvantage for the forecasting model because the effective parameters are not optimized simultaneously and the calculation time is often longer.…”

Data Validity Analysis Based on Reinforcement Learning for Mixed Types of Anomalies Coexistence in Intelligent Connected Vehicle (ICV)

“…Yacong Gao et al developed an LSTM short-term traffic prediction model that can more fully capture recent spatio-temporal features and, thus, proposed a short-term traffic speed prediction method based on multi-temporal traffic flow states before and after. This method can accurately reflect the running state of the road in real time [23]. Traffic flow theory provides a theoretical basis for data anomaly detection through the study of traffic mobility, density, speed, and other indicators, which can help researchers to understand the traffic flow on the road and to make a more accurate judgment on data validity detection.…”

Deep Convolutional Neural Network for Indoor Regional Crowd Flow Prediction

“…Traditional statistics-based methods, such as autoregressive integrated move average (ARIMA), seasonal ARIMA (SARIMA) [3], KARIMA [4] and ARIMAX [5], were proposed in order to predict traffic flow by considering both spatial and temporal features. Although statistics-based methods display simple model structures and are easily explained, they are incapable of considering aspects such as individual randomness and nonlinearity, as well as being inapplicable to large-scale scenarios [6]. These concerns limit the application of statistics-based methods in the field of crowd flow prediction.…”