<![CDATA[The alleviation of traffic accidents is part of Goal 3 of the Sustainable Development Goals (SDGs). However, the lack of access to information on traffic accidents in areas with high traffic accident rates, such as Central Java, makes controlling these cases ineffective. To date, no publication has provided an overview of traffic accident patterns in Central Java. Therefore, this study aims to utilize ensemble learning in traffic accident pattern detection based on online news information extraction. Online news is chosen as an alternative data source because it is fast, open, and informative. The model developed in this research is Indonesian Bidirectional Encoder Representative from Trasnformer (IndoBERT) for Named Entity Recognition (NER) in extracting online news information, which produces an accuracy of 0.9601. Then, the information extraction results will be used to understand traffic accident patterns using Random Forest with an f1-score of 0.7474. This research also proposes a Decision Tree-based surrogate model to improve the interpretability of the Random Forest model with an average accuracy of 0.8995.]]>
Copyrights © 2025
