Driver Classification Using K-Means Clustering of Within-Car Accelerometer Data

Abstract

In this study, driving characteristics of 13 different people on a predetermined route have been analyzed by using the driving characteristics of the drivers and the drivers are classified into 3 groups: calm, normal and aggressive. The data recorded by the acceleration meter sensor and the global positioning (GPS) receiver of a smart phone were analyzed using signal processing methods in the computer environment. Based on the connections between the data, the basic data that reveal the driving characteristics are determined. In the current phase of the study, K-means method was used as the classification method. The classification accuracy was investigated by changing the K value. For experimental data, the most accurate results were obtained as 93.3% for K = 5. This result shows that simple 3-axis accelerometers installed in the cars are sufficient for providing necessary features for classifying driving characteristics using very simple classifiers.