Optimized deep CNN based Prediction model for Injury Severity Risk factor in Automobile Crashes

Abstract

Big Data has become a dominant term in describing the exponential growth, accessibility, availability, and widespread use of information—in structured, semi-structured and unstructured format—in a variety of business context. Of the many areas for application, Big Data Analytics is capable of making a major mark in road safety. Analysis of road accidents, vehicular crashes and inflicted casualties and damage are significant areas of application.  Many researchers have tried to solve issues but still, there are gaps in the road accident severity prediction and finding the contributory factors such as season and time of the accident in which the accident frequently occurred. This leads to the challenges in the field of accident analysis and prediction. The investigation of the risk factors that contribute to the severity of the injury in motorcycle crashes is provoking as a challenge problem across the globe. With its intension, a novel prediction model that predicts the risk factors contributing injury severity in motor vehicle crashes is introduced in this work. Initially, the risk factors contributing injury severity like the accident information, vehicle information, personal information and other information are collected. Further, the prediction process is handled using deep learning model, where optimized Deep Convolutional Neural Network (Deep CNN) is used. In order to enhance the performance of prediction, certain parameters of the convolutional layer, dense layer and dropout layer are fine-tuned by a new Sea Lion updated Dragonfly Algorithm (SL-DA) model, which is the hybridized version of Sea Lion Optimization Algorithm (SLO) and Dragonfly Algorithm (DA) Further, the performance of proposed work will be compared over other state-of-the-art models with respect to positive and negative measures as well.