Comparative Analysis of Machine Learning and Deep Learning Models for Internet of Vehicles Classifications

According to the World Health Organization, the number of deaths due to road accidents keeps increasing every year, and over 90% of them are due to human error. Artificial intelligence can assist in improving drivers’ environmental awareness, contribute to reducing human mistakes, and decrease the number of fatalities. This paper proposes a comparative analysis of the performance of traditional machine learning and deep learning algorithms for a binary image classification problem. This research aims to be a foundation for future research on the Internet of Vehicles in several ways, such as the effectiveness of transfer learning in a limited data environment. Six machine learning algorithms were selected, which are Random Forest, K-Nearest Neighbors (KNN), Stochastic Gradient Descent (SGD), Histogram-Based Gradient Boosting, Support Vector Machines (SVM), and Gradient Boosting, and these algorithms will be benchmarked against the deep learning architectures Feedforward Neural Networks (FNN) and Convolutional Neural Network (CNN). The selected algorithms were tested against two datasets, one containing RGB images and the other greyscale. The modeling demonstrated that the CNN VGG-16 architecture achieved a score ranging from 99.79% to 99.83% for all metrics compared to the traditional machine learning algorithm, which had varying scores ranging from 93.63% to 99.95%.