Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2024
- Goals
- Record
- bf72ead1-cde0-4de7-85c3-228b49215f66
- Published
- 2026-10-30
Universitas Dian Nuswantoro conducted research examining a Single Block Encoder-Decoder Transformer model for multi-step traffic flow forecasting. The activity involved Omar, Mas; Yakub, Fitri; Talip, Mohamad Sofian Abu; Maslan, Mohd Nazmin; Sinha, Vijay Kumar; and Muljono, Muljono as participants. The research addressed the importance of accurate traffic flow forecasting for managing and planning urban transportation systems.
The activity examined a singular horizon-step forecasting method using a proposed Single Block Encoder-Decoder Transformer model optimised with a Grid Search algorithm. It considered the potential of Transformer models in a context where sequence modelling models such as Long Short-Term Memory are widely used. The model was tested on hourly traffic flow data from a single road traffic flow dataset. The data were used to forecast the next 24 hours for the I5-North freeway in California and were sourced from the Caltrans Performance Measurement System.
The optimal configuration included an embedding dimension of 32, a feed-forward dimension of 128, and 8 attention heads. The results showed that the model captured long-range temporal dependencies within the dataset and enhanced forecasting accuracy compared with an LSTM model used for multi-step forecasting. The reported result was a 4.7% reduction in Root Mean Squared Error compared with an LSTM model with two hidden layers of 100 neurons each. This UDINUS research therefore addresses traffic flow forecasting through a Transformer-based method and presents findings concerning urban transportation data and prediction.