Int J Performability Eng ›› 2021, Vol. 17 ›› Issue (10): 880-888.

### An Optimized Intelligent Driver’s Aggressive Behaviour Prediction Model Using GA-LSTM

D. Deva Hemaa,b,* and K. Ashok Kumara

1. aDepartment of Computer Science and Engineering, Sathyabama Institute of Science and Technology, RajivGandhiSalai, Chennai, 600119, India;
bDepartment of Computer Science and Engineering, SRM Institute of Science and Technology, BharathiSalai, Chennai, 600089, India
• Submitted on  ;  Revised on  ; Accepted on
• Contact: *E-mail address: devahema2010@gmail.com
• About author:D. Deva Hema received the M.E Degree in Anna University in 2007 and pursuing Ph.D in the Department of Computer Science & Engineering in Sathyabama Institute of Science and Technology, working as Assistant Professor in SRMIST in Chennai. Her research interests include the area of Artificial Intelligence and Machine Learning including crash prediction and optimization during vehicular crashes.
Dr. K. Ashok kumar received his Master’s Degree in Computer Science and Engineering in 2005 and Ph.D from Sathyabama University, Chennai, Tamilnadu, India in 2016, working as an Associate Professor in the Department of Computer Science & Engineering in Sathyabama Institute of Science & Technology. His research interests are in Grid Computing, Data Analytics and Web Services.

Abstract: Aggressive driving is a significant contributor to traffic accidents. One of the major applications in the area of intelligent transportation systems is identification of aggressive driving behavior. Several deep learning algorithms have been designed for driver behavior prediction, but the unoptimized parameters of Neural Network algorithms suffer to obtain an effective and accurate solution for the prediction of a driver’s aggressive behavior. The unoptimized parameters of aa LSTM model minimize the accuracy and increase the computational cost. Therefore, an intelligent model using the Genetic Algorithm (GA) optimized Long Short Term Memory (LSTM) has been proposed for the prediction of a driver’s aggressive behavior, which maximizes the efficiency of the driver behavior prediction system. The proposed model optimizes the window size of LSTM, total number of hidden layers and hidden units, and learning rate of LSTM model. The result reveals that the proposed driver’s aggressive behavior prediction model achieves an accuracy of 98.2% and outperforms state of art models. The proposed model minimizes the computational cost drastically. The proposed model supports drivers to avoid collisions by presenting alerts to aggressive drivers. In addition, it minimizes the total number of accidents considerably by predicting aggressive behaviors of drivers and presenting alerts before an accident happens.