Home health care routing problem via off-line learning and neural network

Abstract

As the aging problem of China worsens, a new industry named ‘home care’ is rising, aiming to handle the ever-growing demand of elderly cares that nursing homes could no longer handle. In the meantime, smart and autonomous decision-making is facing a new explosion recently due to the rise of Machine Learning. Therefore this paper tends to build an algorithm that could autonomously allocate demands to workers in home care scenario following complex constraints, based on off-line learning with a neural network approximator replacing the action-value function. Experimental results on several large scale instances show that the fully connected network can precisely predict the action-value based on full information of the environment, enabling the agent to make decisions that outperform the baseline model and algorithm. In addition, managerial insights are drawn according to the results given by the neural network-based agent. Conclusions and future research directions are discussed as well.