Affiliations: [a] Department of Computer Science and Engineering, U.I.E.T, Maharshi Dayanand University, India
| [b] College of Computing Science and IT, Teerthanker Mahaveer University, Moradabad, Uttar Pradesh, India
| [c] Guru Jambheshwar University of Science and Technology, Hisar, India
| [d] Institute of Research and Development, Duy Tan University, Danang, Vietnam
| [e] School of Computer Science, Duy Tan University, Danang, Vietnam
Correspondence:
[*]
Corresponding author. Rainu Nandal, Department of Computer Science and Engineering, U.I.E.T, Maharshi Dayanand University, India. E-mail: rainu_nandal@yahoo.com.
Abstract: In recent years, driver behavior analysis plays a vital role to enhance passenger coverage and management resources in the smart transportation system. The real-world environment possesses the driver principles contains a lot of information like driving activities, acceleration, speed, and fuel consumption. In big data analysis, the driver pattern analyses are complex because mining information is not utilized to feature evaluations and classification. In this paper, a new efficient Fuzzy Logical-based driver behavioral pattern analysis has been proposed to offer effective recommendations to the drivers. Primarily, the feature selection can be carried out with the assist of fuzzy logical subset selection. The selected features are then evaluated using frequent pattern information and these measures will be optimized with a multilayer perception model to create behavioral weight. Afterward, the information weights are trained with a test through an optimized spectral neural network. Finally, the neurons are activated by a recurrent neural network to classify the behavioral approach for the superior recommendation. The proposed method will learn the characteristics of driving behaviors and model temporal features automatically without the need for specialized expertise in feature modelling or machine learning techniques. The simulation results manifest that the proposed framework attains better performance with 98.4% of prediction accuracy and 86.8% of precision rate as compared with existing state-of-the-art methods.
