ROAD ACCIDENT SEVERITY PREDICTION USING MACHINE LEARNING | IJORET – Volume 11- Issue 3 | IJORET-V11I3P3

Road accidents are a major public safety concern and cause thousands of injuries and fatalities every year in India. The increasing accident rate highlights the need for analytical systems that can understand accident patterns and predict severity levels. Traditional accident analysis methods mainly provide historical information and fail to identify risk factors or predict accident severity in advance. To address these challenges, this research proposes a machine learning-based accident severity prediction system. The proposed model uses accident-related factors such as weather conditions, road type, vehicle type, time of accident, location, and human behavior to classify accident severity into categories such as Minor, Serious, and Fatal. The dataset is preprocessed and cleaned, followed by feature selection to identify the most influential variables. Machine Learning algorithms like Random Forest, Logistic Regression, XGBoost, and Neural Networks are trained and evaluated. Among them, XGBoost achieves the best performance with high accuracy in predicting serious accident cases. The system effectively identifies accident-prone conditions and patterns, helping authorities and decision-makers implement preventive measures. The results demonstrate that machine learning can significantly improve accident risk prediction and contribute to better road safety planning.

Machine Learning provides an effective approach to predicting road accident severity. By analyzing attributes such as weather, road type, traffic conditions, vehicle type, and human behavior, ML models can accurately classify accident severity levels and identify the major factors contributing to serious accidents. The ability of machine learning algorithms to learn from historical patterns makes them highly suitable for large and complex accident datasets where traditional statistical methods fail to capture deeper relationships. The proposed system achieved high accuracy, especially with the XGBoost model, and successfully identified key accident-prone conditions. This allows authorities and transportation planners to understand accident trends, develop preventive strategies, and implement targeted safety measures. The model also helps in recognizing high-risk locations and conditions, enabling better decision-making for road safety policy and infrastructure development. Furthermore, machine learning-based predictions can support intelligent traffic management systems, assist emergency response units, and contribute to overall public safety improvements. Although the system performs well, certain limitations still exist, such as dependence on data quality, unbalanced severity categories, and the computational complexity of some advanced models. The accuracy of predictions can be affected when real-world accident data is incomplete or inconsistent. Additionally, some models may require higher processing power for training and deployment. Future improvements can include integrating real-time traffic and environmental data, using advanced deep learning architectures for greater precision, and expanding the dataset to include more features such as driver behavior, vehicle sensors, and road infrastructure quality. Overall, the study demonstrates that machine learning is a powerful tool for enhancing road safety analysis. With continuous development and integration of richer datasets, the system has the potential to evolve into a fully automated accident prediction framework capable of supporting smart city initiatives and reducing road-related fatalities in the long run.

[1] A. Sharma and V. Menon, “Accident Severity Prediction using Machine Learning,” IEEE Transactions on Intelligent Transportation Systems, vol. 15, no. 3, pp. 210–218, 2023. [2] R. Kumar, S. Verma, and L. Singh, “Road Accident Analysis and Prediction using Data Mining Techniques,” Elsevier Safety Science, vol. 98, pp. 45–56, 2022. [3] H. Lee and J. Park, “Traffic Accident Severity Analysis using Decision Tree Methods,” Springer Neural Computing & Applications, vol. 34, pp. 789–802, 2023. [4] P. Sharma and K. Gupta, “Logistic Regression-Based Road Accident Prediction,” International Journal of Engineering Research & Technology (IJERT), vol. 11, no. 5, pp. 120–126, 2022. [5] M. Wang and T. Zhao, “Accident Hotspot Detection using K-Means Clustering,” IEEE Access, vol. 10, pp. 43021–43030, 2023. [6] S. Patel, R. Nair, and A. Thomas, “Neural Network-Based Accident Prediction Model,” Elsevier Expert Systems with Applications, vol. 212, pp. 117–129, 2024. [7] J. Johnson and R. Hughes, “Big Data Analytics for Traffic Accident Analysis,” IEEE Transactions on Big Data, vol. 9, no. 2, pp. 544–556, 2023. [8] D. Mehta and S. Chatterjee, “Deep Learning Approaches for Road Accident Prediction,” Elsevier Neurocomputing, vol. 525, pp. 300–312, 2024. [9] K. Zhang and Y. Liu, “XGBoost-Based Road Accident Severity Prediction,” IEEE Access, vol. 11, pp. 11234–11245, 2023. [10] A. Gupta and R. Mehta, “Analysis of Road Accidents using Machine Learning Algorithms,” Elsevier Transportation Research Procedia, vol. 48, pp. 102–110, 2022. [11] S. Reddy and P. Kumar, “Predicting Traffic Accidents using Logistic Regression and Random Forest,” International Journal of Computer Applications, vol. 183, no. 12, pp. 25–30, 2021. [12] L. Chen and M. Wang, “A Comparative Study of Machine Learning Models for Accident Prediction,” Springer Journal of Big Data, vol. 10, pp. 1–15, 2023. [13] D. Sharma and V. Singh, “Road Accident Severity Analysis using Ensemble Learning Techniques,” IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 3, pp. 3456–3465, 2023. [14] P. Roy and S. Das, “Machine Learning Approaches for Road Accident Prediction,” Elsevier Procedia Computer Science, vol. 167, pp. 1350–1359, 2020. [15] J. Kim and H. Lee, “Deep Neural Network-Based Traffic Accident Severity Prediction,” Springer Soft Computing, vol. 26, pp. 5678–5689, 2022. [16] R. Sharma and M. Kulkarni, “Analysis of Road Accident Data using Data Mining Techniques,” International Journal of Data Science and Analytics, vol. 14, pp. 89–101, 2022. [17] Y. Zhou and X. Li, “Gradient Boosting Methods for Traffic Accident Prediction,” IEEE Access, vol. 9, pp. 55678–55689, 2021. [Citation Format]

Submit Your Paper Call for Papers About IJORET Journal Scope Review Process