Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{199708,
        author = {Mohak Nale and pranav Shinde and Mayank Patil and Kunal Patil},
        title = {A Comparative Analysis of Machine Learning Models for Road Accident Severity Prediction},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {14811-14818},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=199708},
        abstract = {accidents are a major worldwide concern, driving to noteworthy misfortune of life, injuries, and financial harm. Exact expectation of mishap seriousness can help specialists take preventive measures, optimize crisis reaction, and improve street security methodologies. This think about presents a comparative examination of multiple machine learning models for anticipating street mischance seriousness using real-world activity mischance information. The dataset utilized in this inquire about comprises of over 800,000 records and incorporates highlights such as climate conditions, road surface, vehicle sort, speed limits, time components, and location characteristics. The information preprocessing phase included dealing with lost values, encoding categorical factors, and addressing course lopsidedness utilizing the Engineered Minority Oversampling Technique(SMOTE). Three machine learning models—Random Woodland, XGBoost, and a Neural Network—were trained and assessed. Furthermore, an gathering show based on delicate voting was implemented to progress prescient execution.
The models were assessed utilizing standard measurements such as precision, precision, recall, F1-score, and ROC-AUC. Among the person models, XGBoost achieved the most noteworthy precision, whereas the outfit demonstrate outflanked all others, achieving a generally precision of around 91.8%. The comes about demonstrate that gathering learning essentially upgrades expectation exactness and robustness. This inquiries about highlights the adequacy of machine learning techniques in street security examination and gives an adaptable arrangement for real-time mischance seriousness forecast frameworks},
        keywords = {},
        month = {April},
        }