A Novel Approach for Predicting wide range of traffic congestion using deep learning Technique

Identifying traffic bottlenecks and devising solutions to the problem requires researchers and practitioners in transportation to understand how congestion in one place impacts the rest of the network. The dynamics of traffic congestion are often modelled using mathematical equations or simulation approaches. Most methods have flaws, such as unreliable assumptions and a time-consuming parameter calibration procedure. For example, ITS and the Internet of Things (IoT) have facilitated the collection of transportation data. This sets off a chain reaction of data-driven investigations into transportation anomalies.Deep learning theory is a potential technique for dealing with big, multidimensional datasets. For this research, deep learning theory will be used to analyze large-scale transportation networks. Recurrent Neural Networks and Restricted Boltzmann Machines are used to model and forecast the evolution of traffic congestion based on the GPS data from a cab. Numerical research was carried out in Delhi, India, to verify the method's efficacy and efficiency. Using a GPU-based parallel computing system, the forecast's accuracy might reach 88% in as little as six minutes. Congestion evolution trends are displayed using a map-based platform to identify weak links for proactive congestion reduction.