Computational Intelligence Based Hybrid Hyperparameter Tuned Prediction Techniques for COVID-19 Epidemiological Data

Dhamodharavadhani, S.; Rathipriya, R.

The aim of this work is to examine the appropriate predictive model for COVID-19 epidemiological data in the Indian population using Computational Intelligence based Hyper parameter Tuned Regression Techniques (CI-HTR). The development of the COVID-19 Epidemiological Data Prediction (EDP) model for India is proposed by using CI-HTR such as Gaussian process regression (GPR) model and shallow neural network models. For that purpose, this analysis uses a time series data set that includes daily cumulative COVID-19 epidemiological data. In order to increase forecast accuracy, hybrid models are constructed by merging Hyper parameter tuned regression models with Non-Linear Auto Regressive (NAR) neural network. Experiential research in this work reveals significant differences between the different types of CI-HTR models for the COVID-19 fatality prediction. The performance metrics like Root Mean Squared Error (RMSE), and the correlation factor called ‘R-Value’, are used to estimate the proposed methodology for COVID-19 dataset. It has been found that the hybrid GPR prediction model is the superior to the other models for the COVID-19 EDP. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.