Smartphone Motion Sensor Data Processing for Driving Characteristics Classification

Abstract

Driving behavior significantly influences road safety. Unsafe driving behaviors, such as driving under the influence, speeding, and using mobile phones, can lead to serious accidents and fatalities. This research aims to observe driving characteristics by utilizing smartphone motion sensor data. The data collection method involved recording the driver’s smartphone motion sensor during trips. The data were then exported from the system for further processing. The main objective of this study is to process the data by creating a classification model with the best performance in handling smartphone motion sensor data. The results of this research are expected to be implementable models to address road safety issues in the future. Additionally, by utilizing driver characteristic detection technology, awareness of safe driving practices can be enhanced. The research methodology used data mining with machine learning classification modeling using random forest (RF), support vector machine (SVM), and decision tree (DT) methods. The test results indicate that the RF model performed the best with an accuracy of 91.22%. Furthermore, this study found that speed was the most influential factor in identifying safe or unsafe driving behavior. The developed classification model shows the potential to improve traffic management efficiency and contribute to safer transportation. By leveraging driver characteristic detection technology, it is hoped that awareness of safe driving practices will increase, leading to a safer road environment.

Keywords

• data mining
• crisp-dm
• driving behavior
• machine learning
• classification