Analysis of Traffic Accident Patterns Using Association Rule Mining

Abstract

This study analyzed the levels of minor, moderate, and severe traffic accidents in the Palembang Police area from 2015 to 2020 using association rule mining and the apriori algorithm. The study established valuable insights into accident trends and contributing factors by leveraging traffic accident data and determining variable relationships. With a minimum support threshold of 0.05 and a confidence value of 0.5, the processed data revealed 349 total incidents, categorized as follows: 58 minor accidents (16.62%), 168 moderate accidents (48.14%), and 123 severe accidents (35.24%). The findings highlight that moderate-level accidents form the majority, underlining the need for targeted interventions in this category. The application of the apriori algorithm facilitated the identification of frequent itemsets and rules that reveal patterns across accident variables, such as road conditions, road functions, accident types, weather conditions, and victim statuses. This study also demonstrated the practicality of the apriori algorithm in analyzing extensive datasets to extract actionable insights. The processed rules can be a foundation for developing predictive models or decision-making tools to mitigate accident risks. For example, analyzing variables at
different accident levels allows policymakers to identify critical factors contributing to accidents, implement tailored safety measures, and prioritize infrastructure improvements. Furthermore, the study emphasizes the potential of data-driven traffic management and accident prevention approaches. By incorporating modern data mining techniques, stakeholders can transition from traditional data recapitulation to predictive analytics, enabling proactive measures for public safety. Future research can build upon this work by integrating real-time data sources, such as IoTbased traffic monitoring systems, to enhance the prediction accuracy and scope of analysis. Further exploration of mid- and low-confidence rules may provide insights into rare but critical patterns, offering a more comprehensive understanding of accident dynamics. Overall, this research is
crucial to leveraging advanced computational methods for public safety and traffic accident reduction, aligning with global efforts to improve road safety and minimize fatalities.