Recognizing Distraction for Assistive Driving by Tracking Body Parts Using Novel Convolutional Neural Network with LGM Classifier Over Random Forest with Improved Accuracy

In the context of identifying driver distractions, the goal of this study is to investigate the degree of accuracy exhibited by the most recent iterations of deep learning algorithms. Both Convolutional Neural Networks Utilising LGM Classifier (CNNLGM) and Random Forest (RF) are compared head-to-head in the research presented here. The investigation required a total of 118 samples, which were then divided equally between two categories consisting of 58 specimens each. Group 1 utilized the CNNLGM Classifier, in contrast to Group 2's utilization of the RF technique. The RF code was implemented using software from Google Colab, and the same program was also used to import the dataset. A pre-test power of 80% and an alpha value of 0.05 were taken into consideration while determining the appropriate size of the sample for the experiment. This was accomplished with the assistance of an online tool for statistical analysis. Previous research provided the necessary information for determining the appropriate size of the sample to use. The findings of the simulation showed that the Novel CNNLGM Classifier achieved an accuracy of 96%, whereas the RF algorithm could only achieve an accuracy of 82%. There was a substantial disparity in the levels of accuracy achieved by the two approaches, as measured by a significance value of 0.001 (p<0.05). To summarise, when utilizing the data that was supplied, the CNNLGM Classifier performed noticeably better than the RF algorithm when it came to identifying when a driver was paying attention.