Artificial neural network-driven approaches to improved forecasting of disability care expenditures in an aging Kingdom of Saudi Arabia population

The total number of older persons globally (those aged 60 years and above) was 202 million in 1950; this total multiplied to attain 901 million and is predicted to triple again in 2100. The growth percentage of the elderly population is quickly improving, and the value of their care shall pose a challenging problem in the future. Notably, the number of older persons in the Kingdom of Saudi Arabia (KSA) is fast growing, from 5% of the entire population in 2015 to a predicted 20.9% by 2050. The main problem is the KSA's management of the rising problem of age-related Non-Communicable Diseases (NCD). With the escalating dimensions of the population of older persons and increased prevalence of NCD causes of risk, the occurrence of NCDs in KSA will rise, resulting in a proportional increase in the requirement for medical assistance. In this paper, an Artificial Neural Network-Based Approaches for Improved Forecasting of Disability Care Expenditures in an Aging Kingdom of Saudi Arabia Population (ANNFDCE-AKSAP) method is proposed. The main objective of the ANNFDCE-AKSAP method is to create an accurate and scalable forecasting system capable of addressing the Kingdom's evolving disability care needs. Initially, the ANNFDCE-AKSAP technique utilizes a min-max normalization-based data preprocessing model to ensure consistent scaling across variables. Furthermore, the bidirectional variational autoencoder with the self-attention module (BiVAE-SAM) model forecasts disability care expenses. Finally, the enriched coati osprey algorithm (ECOA)-based hyperparameter selection process is performed to optimize the prediction results of the BiVAE-SAM method. A wide range of simulations is accomplished to demonstrate the enhanced performance of the ANNFDCE-AKSAP technique, and the results are inspected using several measures. The comparison study of the ANNFDCE-AKSAP technique illustrated the lowest MSE, 0.0128, and MAE, 0.0942, compared to all other methods.