Using Tobit exponential smoothing to enhance demand planning in the presence of censored data
本文提出Tobit ETS模型，扩展了指数平滑法以处理供应链中常见的缺货等删失数据，通过真实和模拟数据验证，显著降低了预测偏差，有助于提升需求规划准确性。
Exponential Smoothing (ETS) is a widely utilised forecasting method in both research and practical applications. A major advancement was the development of a robust statistical foundation through state-space models with a single source of error. However, a significant unresolved challenge is handling censored data, a critical issue in supply chain management, particularly when dealing with stockouts. This paper introduces Tobit ETS, an extension of ETS models capable of effectively addressing censored data scenarios. The proposed model leverages the ETS taxonomy and extends it to encompass censored observations. Results demonstrate a significant reduction in forecast bias, validated through real and simulated datasets from supply chain industries and forecasting competitions. This innovation offers a novel solution for enhancing demand planning accuracy in censored environments.