AI-Based Police Traffic Safety System

Overview

The goal of this project is to develop an innovative service that can predict and prevent traffic accident risks in advance through the systematic collection and analysis of traffic data. The core objective is to establish an AI-based traffic accident risk prediction model utilizing diverse public and private sector data, and to develop a nationally scalable traffic safety system by conducting field tests in collaboration with the Korean National Police Agency.

South Korea maintains a high level of traffic fatalities among OECD countries, and the social and economic losses from traffic accidents persist. Existing traffic safety systems focus primarily on responding after accidents occur, increasing the need for preventive traffic safety services.

Accordingly, there is a need to establish a proactive response system based on data and to develop services that predict the likelihood of traffic accidents using AI technology.

Project Description

This project trained an AI model using traffic accident data held by the National Police Agency and the Korea Road Traffic Authority, establishing a system to predict traffic accident risk levels. Specifically, it analyzed the causes of traffic accidents by utilizing 11 types of data, including traffic enforcement data, road facility information, weather data, and traffic volume data.

The data lake-based infrastructure was used for data analysis and AI model development, establishing a data lake system operable within the Korean National Police Agency. Efficiency was increased by applying automated infrastructure (MLOps) from data collection through preprocessing, analysis, and model training.

Additionally, a traffic accident risk prediction model was developed to assess accident risk in specific areas, followed by field validation in collaboration with the National Police Agency. The system was deployed in high-risk regions—Pyeongtaek, Chungju, and Yeongam—where its impact on reducing traffic accidents was evaluated.

Key technological features

• Development of Traffic Accident Risk Prediction Models Using Machine Learning and Data Mining Technologies
• Core Algorithms: Random Forest, LightGBM, AdaBoost
• Leveraging Data Lake: Automating Data Collection and Processing, and Establishing an MLOps Environment

Expected outcomes

The key achievements of this project include the establishment of an AI-based traffic accident risk prediction system in collaboration with the National Police Agency and the validation of its effectiveness through real-world demonstrations. This has laid the groundwork for proactively managing high-risk areas with elevated accident likelihood. By integrating diverse traffic datasets and applying AI-driven analysis, the project introduced a predictive and preventive model that improves upon conventional, post-incident traffic safety approaches.

When linked with future traffic accident prevention policies, the system is expected to further reduce traffic-related fatalities, lower social costs, and enhance the overall safety of road users. Going forward, additional traffic data will be incorporated, and real-time data will be utilized to further improve model accuracy. To this end, the project plans to collaborate with private companies to integrate real-time traffic information. Continuous model updates and performance optimization will also be carried out to ensure stable and sustainable system operation within the National Police Agency.

Launch year: 2024

Attribution

Case study courtesy of the National Information Society Agency, Republic of Korea, Apolitical's Content Partner. Sourced from Use Cases of Public AI Service, Vol.1,  https://eng.nia.or.kr/site/nia_eng/ex/bbs/View.do?cbIdx=31975&bcIdx=28997&parentSeq=29045