Medical waste generation in South Korea rose through 2018 but has since declined structurally, owing to the 2019 reclassification of disposable diapers as general waste, the COVID-19 pandemic, and the 2024-2025 resident-doctor strike, even as population aging drives underlying demand upward. This study couples demand forecasting with linear programming (LP)-based policy evaluation. An ensemble of ARIMAX (autoregressive integrated moving average with exogenous variables) and Ridge regression, trained on 2000-2024 national data from the Korean Statistical Information Service (KOSIS), attains a holdout mean absolute percentage error (MAPE) of 14.42% and forecasts that generation will surpass the current treatment capacity of 259,150 tonnes per year around 2026. Three policy scenarios are evaluated over 2025-2034 using a regional multi-period LP with interregional transport and facility lead times: (S1) on-site sterilization expansion, (S2) new incineration, and (S3) expansion of existing incineration. On-site expansion coincides with the unrestricted optimum under Low and Medium demand, yielding a net present cost about 8.3% and 9.9% below existing-plant and new-plant incineration, respectively; it remains cheaper once the adoption ceiling exceeds roughly 25% even under High demand, and new incineration is never optimal. These findings support the recent regulatory relaxation of on-site sterilization standards and indicate that diversifying away from centralized incineration is economically viable.

1. 서 론1
2. 기존 연구
    2.1 의료폐기물 발생량 예측 관련 연구
    2.2 의료폐기물 처리시설 입지 및 용량 최적화연구
    2.3 기존 연구의 한계 및 본 연구의 차별성
3. 의료폐기물 발생량 예측
    3.1 데이터 수집 및 전처리
    3.2 모형 성능 비교 및 검증
    3.3 확장윈도우 홀드아웃 교차검증
    3.4 중장기 의료폐기물 발생량 예측 결과
4. 의료폐기물 대응 전략 시나리오 및 모형 설계
    4.1 정책 시나리오
    4.2 집합, 변수, 파라미터
    4.3 모형 정식화
    4.4 수요 시나리오
    4.5 주요 가정
5. 결과 분석
    5.1 파라미터 설정
    5.2 시나리오별 결과
    5.3 도입 수준 민감도와 조건부 결론
6. 결 론
References

• Hyebeen Jung(Department of Industrial and Data Engineering, Hongik University) | 정혜빈 (홍익대학교 공과대학 산업․데이터공학과)
• Chunghun Ha(Department of Industrial and Data Engineering, Hongik University) | 하정훈 (홍익대학교 공과대학 산업․데이터공학과) Corresponding author