Transportation System Using Deep Learning Algorithms in Industry 4.0 Towards Society 5.0

Late years have seen a lot of transportation information gathered from various sources including street sensors, tests, GPS, CCTV, and factual reports. Like numerous different businesses, transportation has entered the age of large information. With a rich volume of traffic records, it is trying to assemble a dependable estimation methodology that is dependent on conventional trivial Machine Learning (ML) technique. AI plays the center capacity to intellectualize the transportation frameworks. Late years have seen the approach and pervasiveness of profound realizing which has incited a detailed study in smart transportation framework. Therefore, conventional ML models in the number of applications have been implemented by the new learning strategies. A thorough literature survey of deep learning algorithms is being done which can be implemented for enhancing the transportation system. In this chapter, we have discussed various deep learning models that are specifically used for the prediction of traffic flow. The utilization of Deep Learning (DL) frameworks in transportation is yet to be explored more in detail and there is a lot of problems for DL models to be implemented in the transportation system. By combining non-linear modules, the transformation of representation from one level to another higher and abstract level is deep learning algorithm. There are sufficient numbers of transformations which will help in learning the most complicated functions and structures for efficiently. The main advantage of DL is selection of the features by using general purpose learning methodology without any human involvement. DL methodologies has given result in the demonstration of high performance by exploring the high-dimensional data in several domains like computer vision, natural language processing, bioinformatics, etc. There are numbers of DL methodologies consisting of Recurrent Neural Networks (RNNs), Deep Convolutional Networks (DCNs), Deep Restricted Boltzmann Machines (RBM), Stacked Auto-Encoders (SAEs), Deep Belief Networks (DBN), etc.