Abstract

In this paper, we present the design and development of an Internet of Things (IoT)-integrated car rental system that collects and monitors data remotely in real time using an OBD-II interface and a mobile application. The proposed system gathers critical information about both vehicles and users, including driver identity, trip time, driving speed, fuel consumption, and vehicle location. All data is transmitted through an IoT platform and stored on a cloud-based mobile application, enabling easy and transparent access for both users and service providers.

Beyond vehicle monitoring, the system integrates an optimized AdaBoost algorithm using GridSearchCV to evaluate driving behavior. This model analyzes data collected from the OBD-II interface to identify unsafe driving behaviors, such as frequent sudden braking, abrupt acceleration, and unstable speed control.

Experimental results reveal significant differences in the performance of the evaluated machine learning algorithms. Decision tree and logistic regression models yielded the lowest results, with accuracy values of approximately 69% and 68%, respectively, and their precision, recall, and F1-score metrics remained below 70%. In contrast, the random forest, support vector machine, and k-nearest neighbors models achieved better and relatively similar performances, with accuracy values around 73%. Among all models, our optimized AdaBoost algorithm delivered the best classification results, with accuracy, precision, recall, and F1-score all reaching approximately 80%.

Based on observed behaviors, driver profiles will be created and stored. These profiles can enhance service quality, support risk management, and enable more personalized rental policies. Test results confirm that the system is feasible for real-time operations and provides valuable insights into driver behavior.