Relief demand prediction and operational planning in flood-affected areas
研究提出一个预测与决策框架，结合时间序列和机器学习方法预测印度洪水受灾地区的救援需求，并链接到运营规划，帮助人道组织优化预算和资源分配。
Natural disasters result in loss of human lives, economic damage, and disruption to living conditions. In one way, the losses can be minimised by better planning of post-disaster relief operations, such as forecasting relief demands and allocating resources to the affected population. This study proposes a forecasting and decision-support framework to predict relief demand in flood-affected regions, enabling effective operational planning at both the national and state levels in India. The time series and machine learning methods were explored for forecasting, and their performance was evaluated using error metrics. Furthermore, a forecasting framework that combines principal component analysis and autoregressive integrated moving average models with Random Forest methods has been proposed, based on the observation from state-of-the-art methods and flood data characteristics. Furthermore, this study links forecasting to operational planning, in which the requirements for various items have been quantified, and the annual demand quantities disaggregated into pre-monsoon, main-monsoon, and post-monsoon flood waves. Overall, this study provides an integrated forecasting and operational decision-making framework that humanitarian organisations can adopt to support effective operational planning and budget allocation under the Total Cost of Ownership approach.