The semi-supervised classification of petrol and diesel passenger cars based on OBD and support vector machine algorithm

Abstract

This paper proposes a novel semi-supervised classification method of petrol and diesel passenger cars using OBD data and support vector machine (SVM) algorithm. The proposed method first develops a classification rule of petrol and diesel passenger cars based on vehicle speed as well as engine RPM obtained from the on-board diagnostic (OBD) data with specific passenger car identification number (ID). Then the proposed method could primarily label petrol or diesel to the passenger car with specific ID using the classification rule. Next this paper apply support vector machine to create a classification model of petrol and diesel passenger cars based on the primary classification results, and to perform refined classification tasks. Experimental results show the correctness of the proposed semi-supervised petrol and diesel passenger car classification method can achieve 1.5% calibration rate from more than 35,000 real OBD data. The proposed method has the potential of applying to internet of vehicle (IoV) and to improve on-road CO2 emission estimation.