基于LSTM-GA混合模型的患者预约排队策略优化

在医疗运作管理领域,合理的资源分配能够帮助更多的患者尽早就医,降低患者病情恶化和死亡的风险。本文设计了预约排队策略对患者占有资源的顺序进行分配,建立了基于长短时记忆(Long Short Term-Memory, LSTM)神经网络和遗传算法(Genetic Algorithm, GA)的混合模型以优化排队策略。首先利用大数据和深度学习分析患者到达和医院服务情况,建立LSTM神经网络学习数据特征并预测未来数据,相比于排队论常用的随机分布方法取得了更好的效果.其次设计了基于排队系统仿真的排队策略优化算法,利用改进GA得到最优排队策略。实证研究表明,文本的方法可以明显降低患者的等待时间,最高可达59%。最后对排队策略进行敏感性分析,结果表明排队策略有效作用于仿真的各个时段。

Optimization of Patient Reservation Queuing Policy Based on LSTM-GA Hybrid Model

In the field of medical operation and management, a reasonable allocation of resources is conducive to more patients to seek medical treatment as soon as possible and reduce the risk of patients' disease deterioration and death. Reservation queuing policy is designed to allocate the order of patients' occupying resources. A hybrid model based on long Short Term Memory(LSTM)neural network and genetic algorithm(GA)is established to optimize the queuing policy. First, big data and deep learning are used to analyze patient arrival and hospital service, and LSTM neural network is established to learn data features and predict future data. Compared with the random distribution method commonly used in queuing theory, it has achieved better results. Secondly, a queuing policy optimization algorithm based on queuing system simulation is designed to obtain the optimal queuing policy by using improved GA. The empirical studies show that the method of text can significantly reduce the patient's waiting time, which lowers the target function by 59%. Finally, the sensitivity of the queuing policy is analyzed, and the results show that the queuing policy is effective in each period of the simulation.