Covid-19 Daily Case Modelling using Nonparametric Regression Hybrid Model

Abstract

Nonparametric Regression Model is used to analyze data patterns and predict the future based on past data. The problem of Nonparametric Regression research so far has only used one approach, causing the estimation results to be biased, even though each sub-pattern of data has its own suitability depending on the approach method used. In this study, a hybrid nonparametric regression modeling model based on two approaches will be developed. The approach used in this research is Local Polynomial and Fourier Series. Coronaviruses are part of the family of viruses caused by the corona virus, otherwise known as COVID-19, which is a new type that was discovered in 2019. The spread of the corona virus in Semarang City continues to increase dramatically. Therefore, the purpose of this study is to estimate the Local Polynomial- Fourier Series Hybrid model which is
then applied to the Covid-19 case in Semarang City. The following is the research plan or research stages for the Primary Research Scheme, namely estimating the Hybrid model with two Local Polynomials and Fourier Series approaches, the estimation method used is Weighted Least Square
(WLS). In the Local Polynomial- Fourier Series hybrid model there is a smoothing parameter selection, the method used is the Generalized Cross Validation (GCV) method. Furthermore, the program that has been created is applied to Covid-19 data in Central Java so as to produce Covid19
case modeling based on Hybrid Nonparametric Regression using two approaches, namely Local Polynomials and Fourier Series. The processed
results using nonparametric regression hybrid model of local polynomial
- fourier series approach applied to daily case data of Novel Coronavirus in
Semarang City showed a coefficient of determination of 0.9968 (99.68%) and an MSE of 313.153.

Keywords: Covid-19, Model Hybrid, Local Polynomial, Nonparametric Regression