Optimization of Driver Behavior Profiling through K-means Unsupervised Clustering Algorithm Using Real-world Data

Abstract

The use of telematics systems has brought benefits as conventional modern cars are now equipped with the technology, allowing data gathering to a centralized system, allowing vehicle monitoring and management. The data gathered from the sensors can provide insight into driver behavior and driving patterns. Unfortunately, the data is not fully utilized for in-depth analysis, as the pattern can be too
sophisticated to understand. This hinders its potential to improve the safety of the road environment, as providing information from the data pattern can be utilized by drivers, law enforcement, policy makers, or anyone related to road safety management. This paper provides an in-depth analysis of real-world driving data behavior using unsupervised algorithms (K-means). The paper aims to assess the data pattern of the driver. The study uses the K-means algorithm to cluster the data of drivers and separate it by pattern. Further analysis is needed to classify driving based on characteristics of the clusters, such as bad and good drivers. The findings reveal that K-means was able to identify patterns of the driving behavior, and analysis was done for categorization. 6 clusters were identified in the algorithm, where clusters 2 and 4 exhibit good driving patterns while clusters 1,3,4, and 6 exhibit bad driving patterns. This research provides crucial information to the driver’s awareness, gives insight into policymakers and law enforcement, thus improving the safety of the road.