The Use of Machine Learning Methods in Road Safety Research in Poland

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(2), P. 861 - 861

Published: Jan. 16, 2025

Every year, thousands of accidents occur in Poland, often resulting severe injuries or even death. The implementation solutions supporting road safety analysis and management processes is necessary to reduce the risk minimize their consequences. One rapidly developing tools that can play a key role this area machine learning. aim study was develop mathematical models based on ML algorithms describing Poland. First, variables with strongest impact were extracted. Then, modeling performed using k-Nearest Neighbors, Random Forest, RPart algorithms. best choice for imbalanced data, especially when goal identify rare classes, RF model. KNN model provides compromise situations where highest overall accuracy desired. On other hand, be used as fast, basic model, but it requires improvements handle classes. results not only identified factors significantly affect severity number fatalities but, above all, also demonstrated ability ML-based predict threats

Language: Английский

Ordinal Random Tree with Rank-Oriented Feature Selection (ORT-ROFS): A Novel Approach for the Prediction of Road Traffic Accident Severity

Mathematics, Journal Year: 2025, Volume and Issue: 13(2), P. 310 - 310

Published: Jan. 18, 2025

Road traffic accident severity prediction is crucial for implementing effective safety measures and proactive management strategies. Existing methods often treat this as a nominal classification problem use traditional feature selection techniques. However, ordinal that account the ordered nature of (e.g., slight < serious fatal injuries) in still need to be investigated thoroughly. In study, we propose novel approach, Ordinal Random Tree with Rank-Oriented Feature Selection (ORT-ROFS), which utilizes inherent ordering class labels both stages classification. The proposed approach enhances model performance by separately determining importance based on levels. experiments demonstrated effectiveness ORT-ROFS an accuracy 87.19%. According results, method improved 10.81% over state-of-the-art studies average different train–test split ratios. addition, it achieved improvement 4.58% methods. These findings suggest promising accurate prediction, supporting road planning intervention

Language: Английский

Unbalanced Traffic Accident Severity Prediction: Fusing Vae-Attention and Gcn Algorithm

Published: Jan. 1, 2025

Language: Английский