PROJECT TITLE :

A Multitask Learning Model for Traffic Flow and Speed Forecasting

ABSTRACT:

Accurate short-term traffic state forecasting is beneficial to Intelligent Transportation Systems (ITS) research and applications.

This research offers a deep learning-based multitask learning Gated Recurrent Units (MTL-GRU) with residual mappings to improve forecasting accuracy. Feature engineering is used to identify the most informative features for forecasting in order to improve the MTL-performance.

GRU's The MTL-GRU can then estimate traffic flow and speed simultaneously, and performs better than other analogues, according to numerical results based on real-world datasets. Experiments also show that the MTL-GRU model, which is based on deep learning, can overcome the bottleneck created by larger training datasets while still gaining benefits.

The findings indicate that the suggested MTL-GRU model with residual mappings is capable of forecasting short-term traffic conditions.