This paper aims to enhance the effectiveness of emergency response in the event of infectious disease outbreaks, reduce the population infection rate, contain the scope of epidemic spread and minimise losses from casualties.
This study pioneers the examination of emergency supply allocation in epidemic regions amid public health crises, formulating a multi-objective optimisation model. The model addresses two types of locations – demand points and distribution centres – and three categories of emergency supplies: food, daily necessities and medical supplies. To address the issue of uncertainty in demand for emergency materials, this paper compares evaluation indicators and performs an error analysis based on emergency material demand forecasts using long short-term memory (LSTM) and susceptible–exposed–infectious–recovered–susceptible (SEIRS) models.
The empirical results demonstrate that the LSTM model significantly outperforms the SEIRS model in terms of forecast accuracy. Given this, the study employs LSTM networks to extract time-series features from real-time epidemic information, enabling dynamic assessment of emergency material demand and real-time prediction of demand in each epidemic area. We develop a novel Particle Swarm Optimization–Gravitational Search Algorithm (PSO-GSA), integrating particle swarm optimisation and gravitational search algorithms, to solve this complex model. Using Hubei Province's 202 epidemic data as a case study, we aim to minimise distribution time, unmet demand and total distribution cost.
This paper establishes a dynamic demand prediction model of LSTM, and applies LSTM to emergency material demand prediction in public health emergencies for the first time. This paper improves the algorithm and optimises the parameters, and proposes a new hybrid algorithm, PSO-GSA. Through comparative analysis, it is proven that LSTM prediction and the PSO-GSA algorithm have significant advantages in practical scenarios, which provide a feasible decision support for the dynamic deployment of emergency supplies during public health emergencies.
A diagram of a multi-objective optimization model for emergency materials distribution. The diagram features a central distribution center depicted as a warehouse with a truck, surrounded by four requirement points represented by clusters of buildings with red virus icons. Arrows indicate the flow of materials from the distribution center to the requirement points. The top section of the diagram includes text boxes describing different algorithms and their applications. The left text box mentions LSTM and SEIRS, highlighting the reduction in prediction errors. The middle text box introduces a hybrid PSO-GSA algorithm combined with LSTM for forecasting demand during public health emergencies. The right text box compares the hybrid PSO-GSA algorithm with traditional PSO algorithm and GA. The overall structure emphasizes the dynamic extraction of real-time epidemic information to optimize the distribution of emergency materials.A diagram of a multi-objective optimization model for emergency materials distribution. The diagram features a central distribution center depicted as a warehouse with a truck, surrounded by four requirement points represented by clusters of buildings with red virus icons. Arrows indicate the flow of materials from the distribution center to the requirement points. The top section of the diagram includes text boxes describing different algorithms and their applications. The left text box mentions LSTM and SEIRS, highlighting the reduction in prediction errors. The middle text box introduces a hybrid PSO-GSA algorithm combined with LSTM for forecasting demand during public health emergencies. The right text box compares the hybrid PSO-GSA algorithm with traditional PSO algorithm and GA. The overall structure emphasizes the dynamic extraction of real-time epidemic information to optimize the distribution of emergency materials.
