The application of interpretable machine learning model based on comparative learning and NARMAX in epidemic research
Sibei Cheng, Jin Zhou, Xiao Qing Luo, Tao Zhou, Bochou Jiang · 2022 IEEE International Conference on Artificial Intelligence and Computer Applications (ICAICA) · 2022
Most of the machine learning models are black box models. However, in practical applications, such as in many medical and health fields, it is very necessary to clearly understand the internal composition, combination or interaction of the model, study the system and predict the system behavior. Therefore, interpretable machine learning models have attracted more and more attention, especially when predicting based on models, the driving factors leading to prediction behavior are deeply studied. This paper proposes an interpretable machine learning model based on comparative learning and NARMAX. Because the input-output relationship of the model and the interaction relationship between input variables are clear, the model can not only be used for prediction, but also explain the relevant “reasons” of prediction behavior. The novel coronavirus pneumonia epidemic data and influenza epidemic data were used to compare the model proposed in this paper. The experimental results show that the model is effective and reliable, and establish a dynamic model for the two diseases’ spreads, and analyze the relationship between disease transmission factors.