랜덤 포레스트와 패널 로지스틱 모델을 활용한 양돈농장의 아프리카돼지열병(ASF) 발생위험 예측: 시군구 단위 분석 = Predicting the Risk of African Swine Fever (ASF) Occurrence in Pig Farms Using Random Forest and Panel Logistic Models: A District-Level Analysis in South Korea

다국어 초록 (Multilingual Abstract)

This study aims to provide foundational information to support the development of effective farm- and district-level (si-gun-gu) biosecurity policies against African Swine Fever (ASF) outbreaks in South Korea. Using ASF outbreak data and detailed farm characteristics from 2019, when ASF was first detected in Korea, to April 2023, we estimated farm-level ASF occurrence probabilities through random forest and panel logistic regression models. These probabilities were then aggregated to the district level to facilitate practical biosecurity management. Key factors significantly influencing ASF risk included farm distances from the nearest farm, distancesfrom the nearest ASF-infected farm, herd size, number of neighboring farms within 5 km and 10 km radii, frequency of ASF-positive wild boar detections within a 4 km radius, number of ASF-infected farms within the district, farm distances from roads, and land-use characteristics such as areas of cultivated land, forest, and rivers. Both prediction models demonstrated high accuracy; the random forest model was particularly effective in short-term forecasting, whereas the panel logistic model demonstrated greater utility for medium- to long-term forecasting. Further studies are necessary to clarify the precise causal mechanisms through which these factors influence ASF occurrence probabilities.