A hybrid FSRF model based on regression algorithm for diabetes medical expense prediction.

The number of patients with diabetes continues to grow, and the expense of treating diabetes is enormous. Therefore, predicting medical expenses for diabetes has become a priority in many countries. This paper proposes a new hybrid FSRF model to predict medical expenses. Firstly, in response to the problem of multiple features in medical data, we use a random forest (RF) feature extraction algorithm for feature extraction. Secondly, for complex medical concepts, we develop an improved multi-granularity embedding model for encoding medical concepts. Next, we establish the FA-SSA by optimizing the sparrow search algorithm (SSA) using the firefly algorithm (FA). Then, we employ the FA-SSA algorithm to optimize the parameters of the RF model with multi-granularity medical concept embedding. Finally, we build an improved FSRF model and conduct a case study on a medical dataset in Pingjiang County. This paper performs ablation experiments and four sets of comparative experiments, and the experimental results show the superiority of the FSRF model. • The RF algorithm is introduced to address the problem of massive redundancy in medical features. • A multi-granularity medical concept embedding model is developed for medical concept coding. • An FA-SSA parameter optimization algorithm is proposed. • An improved FSRF model is proposed for medical cost prediction studies. • Validate the model using medical data from Pingjiang County. [ABSTRACT FROM AUTHOR]