Road Type Classification of Driving Data Using Neural Networks

Road classification, knowing whether we are driving in the city, in rural areas, or on the highway, can improve the performance of modern driver assistance systems and contribute to understanding driving habits. This study focuses on solving this problem universally using only vehicle speed data. A...

Full description

Saved in:
Bibliographic Details
Main Authors: Dávid Tollner, Máté Zöldy
Format: Article
Language:English
Published: MDPI AG 2025-02-01
Series:Computers
Subjects:
Online Access:https://www.mdpi.com/2073-431X/14/2/70
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849719668774797312
author Dávid Tollner
Máté Zöldy
author_facet Dávid Tollner
Máté Zöldy
author_sort Dávid Tollner
collection DOAJ
description Road classification, knowing whether we are driving in the city, in rural areas, or on the highway, can improve the performance of modern driver assistance systems and contribute to understanding driving habits. This study focuses on solving this problem universally using only vehicle speed data. A data logging method has been developed to assign labels to the On-board Diagnostics data. Preprocessing methods have been introduced to solve different time steps and driving lengths. A state-of-the-art conventional method was implemented as a benchmark, achieving 89.9% accuracy on our dataset. Our proposed method is a neural network-based model with an accuracy of 93% and 1.8% Type I error. As the misclassifications are not symmetric in this problem, loss function weighting has been introduced. However, this technique reduced the accuracy, so cross-validation was used to use as much data as possible during the training. Combining the two approaches resulted in a model with an accuracy of 96.21% and unwanted Type I misclassifications below 1%.
format Article
id doaj-art-91ff509b4dd24d5eb495edbfbccf5ce2
institution DOAJ
issn 2073-431X
language English
publishDate 2025-02-01
publisher MDPI AG
record_format Article
series Computers
spelling doaj-art-91ff509b4dd24d5eb495edbfbccf5ce22025-08-20T03:12:05ZengMDPI AGComputers2073-431X2025-02-011427010.3390/computers14020070Road Type Classification of Driving Data Using Neural NetworksDávid Tollner0Máté Zöldy1Department of Automotive Technologies, Budapest University of Technology and Economics, 1111 Budapest, HungaryDepartment of Automotive Technologies, Budapest University of Technology and Economics, 1111 Budapest, HungaryRoad classification, knowing whether we are driving in the city, in rural areas, or on the highway, can improve the performance of modern driver assistance systems and contribute to understanding driving habits. This study focuses on solving this problem universally using only vehicle speed data. A data logging method has been developed to assign labels to the On-board Diagnostics data. Preprocessing methods have been introduced to solve different time steps and driving lengths. A state-of-the-art conventional method was implemented as a benchmark, achieving 89.9% accuracy on our dataset. Our proposed method is a neural network-based model with an accuracy of 93% and 1.8% Type I error. As the misclassifications are not symmetric in this problem, loss function weighting has been introduced. However, this technique reduced the accuracy, so cross-validation was used to use as much data as possible during the training. Combining the two approaches resulted in a model with an accuracy of 96.21% and unwanted Type I misclassifications below 1%.https://www.mdpi.com/2073-431X/14/2/70CAN bus dataon-board diagnosticsdata loggingroad type classificationneural network
spellingShingle Dávid Tollner
Máté Zöldy
Road Type Classification of Driving Data Using Neural Networks
Computers
CAN bus data
on-board diagnostics
data logging
road type classification
neural network
title Road Type Classification of Driving Data Using Neural Networks
title_full Road Type Classification of Driving Data Using Neural Networks
title_fullStr Road Type Classification of Driving Data Using Neural Networks
title_full_unstemmed Road Type Classification of Driving Data Using Neural Networks
title_short Road Type Classification of Driving Data Using Neural Networks
title_sort road type classification of driving data using neural networks
topic CAN bus data
on-board diagnostics
data logging
road type classification
neural network
url https://www.mdpi.com/2073-431X/14/2/70
work_keys_str_mv AT davidtollner roadtypeclassificationofdrivingdatausingneuralnetworks
AT matezoldy roadtypeclassificationofdrivingdatausingneuralnetworks