Driver distraction detection using experimental methods and machine learning algorithms.
| dc.contributor.advisor | Velenis, Efstathios | |
| dc.contributor.advisor | Fotouhi, Abbas | |
| dc.contributor.advisor | Auger, Daniel J. | |
| dc.contributor.author | Zhang, Zhaozhong | |
| dc.date.accessioned | 2024-03-07T13:31:54Z | |
| dc.date.available | 2024-03-07T13:31:54Z | |
| dc.date.issued | 2020-02 | |
| dc.description.abstract | Driver distraction causes numerous road accidents, which is approximately equal to 25% of the total crashes according to the reports by the National Highway Traffic Safety Administration. Warnings can be helpful to mitigate the risks caused by driver distraction. Previous studies on driver distraction detection have not sufficiently found relevant input features to filter insignificant information, thus limiting the improvement of efficiency. Moreover, the disadvantages of driving simulators and public roads pose a challenge in collecting suitable data for feature identification and comparisons of performance among driver distraction detection algorithms. While the previous research focuses on improving prediction accuracy, shortening the prediction time is critical in giving timely warnings to drivers. This thesis aims at detecting driver distraction, which could provide faster and accurate warnings to drivers. The developed method is implemented by cutting the redundancy and irrelevant information fed to the algorithms and instead selecting suitable algorithms that achieve the balance between the prediction accuracy and prediction time. Moreover, a closed testing field supplies an environment for collecting more accurate information to identify the relevant features and to determine suitable algorithms. In this study, open-source data and experimental data are used. The results show that a balance between the prediction accuracy and the prediction time is achieved by feeding the relevant features and using suitable machine learning algorithms (e.g. Decision Tree). Compared with existing state-of-the-art methods, the prediction accuracy of the method proposed in this study has reached approximately the same level. More importantly, the efficiency has improved, including reduced prediction time and fewer input features. Consequently, less computer storage is used. | en_UK |
| dc.description.coursename | PhD in Transport Systems | en_UK |
| dc.identifier.uri | https://dspace.lib.cranfield.ac.uk/handle/1826/20947 | |
| dc.language.iso | en | en_UK |
| dc.publisher | Cranfield University | en_UK |
| dc.publisher.department | SATM | en_UK |
| dc.rights | © Cranfield University, 2020. All rights reserved. No part of this publication may be reproduced without the written permission of the copyright holder. | en_UK |
| dc.subject | driver distraction | en_UK |
| dc.subject | data visualisation | en_UK |
| dc.subject | decision tree | en_UK |
| dc.subject | logistic regression | en_UK |
| dc.subject | machine learning | en_UK |
| dc.subject | relevancy analysis | en_UK |
| dc.subject | visual inspection | en_UK |
| dc.title | Driver distraction detection using experimental methods and machine learning algorithms. | en_UK |
| dc.type | Thesis or dissertation | en_UK |
| dc.type.qualificationlevel | Doctoral | en_UK |
| dc.type.qualificationname | PhD | en_UK |