Driver behavior classification and its impact on fuel consumption using machine learning techniques: A comprehensive literature review

• Reviews machine learning methods linking driver behavior and fuel consumption. • Identifies key gaps in data standards, generalization, and sensing. • Compares supervised and unsupervised models for behavior detection. • Maps behavior classes to fuel outcomes to inform eco-driving strategies. • Outlines a roadmap with standardized datasets and scalable ML pipelines. With the growing number of vehicles globally, numerous challenges have emerged, including a rise in trip frequency, leading to an increase in accidents and heavy traffic in both urban and rural areas, ultimately affecting fuel consumption. Recent research highlights that driver behavior plays a critical role in these issues. The primary objective of this study is to provide a comprehensive literature review on the classification of driver behavior using Machine Learning techniques. Key research questions include how machine learning can support the classification of driving behavior and its correlation with fuel consumption. This study employs a bibliometric review methodology following the PRISMA guidelines using the Scopus database, analyzing publications from January 2012 to April 2024. Bibliometric analysis, keyword mapping, and citation tracking tools were used to identify trends and research gaps. Findings indicate that while supervised algorithms like Random Forest and SVM are most effective for classification accuracy, unsupervised methods are essential for revealing latent driving styles in massive datasets. Crucially, the review reveals that integrating these classification models into real-time feedback systems can lead to substantial reductions in fuel consumption and contribute significantly to environmental sustainability. The study contributes to the field by synthesizing current knowledge, outlining limitations, and recommending future research directions aimed at enhancing sustainable mobility through intelligent behavior modeling.