Prediction of Accident Severity Using Machine Learning Algorithms

Shah, Dhruv Vimal (2023) Prediction of Accident Severity Using Machine Learning Algorithms. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (3MB) | Preview
Preview	PDF (Configuration manual) Download (2MB) | Preview

Abstract

Vehicle accidents are among the most terrifying experiences a person can have, and at times leaves a lifelong mark on the victim. Accidents and collisions are a regular phenomenon, that keeps happening frequently as a direct result of the recklessness of drivers, road conditions, and other environmental conditions. Using crash severity prediction models, various government agencies can get insights into the variables that influence the incidence, allowing them to forecast the severity of an accident. With the aid of accident data, machine learning algorithms can help to find patterns that might help predict the severity of an accident, like fatalities, serious injuries, or just minor injuries. The fundamental purpose of the research was to provide a way to use machine learning algorithms to predict the level of damage caused by an accident. In this study, we made a prediction framework and used three different machine learning algorithms—random forest, logistic regression, and decision tree—to figure out how bad the accident’s impact could be. This study carried out three experiments using a publicly available dataset collected from the Kaggle repository, originally released by the UK Department of Transport. Each of the algorithms was fine-tuned with hyperparameters to boost the classification prediction to gain the best possible results for the study. The random forest model was 86.23% accurate, the logistics regression model was 85.60% accurate, and the decision tree model was 86.23% accurate. The results demonstrate that Random Forest and Decision Tree were the best algorithms in terms of accurately predicting all three accident severity classes, as opposed to Logistic Regression, which only predicted the third class. Following the construction of the models, the results of the experiments were analysed with performance metrics such as accuracy, precision, recall, f1 score, and confusion matrix. With the use of random forests and decision tree algorithms, the proposed solution will help improve road safety and help the authorities in charge of road maintenance come up with plans for reducing accidents.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Ul Ain, Qurrat UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > TE Highway engineering. Roads and pavements T Technology > TL Motor vehicles. Aeronautics. Astronautics Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Tamara Malone
Date Deposited:	25 May 2023 16:33
Last Modified:	25 May 2023 16:33
URI:	https://norma.ncirl.ie/id/eprint/6653

Actions (login required)

View Item