Assessing the significance of socioeconomic and demographic factors on COVID-19 cases in Turkey along with the development levels of provinces

Abstract

In this study, we examine the spatial analysis of coronavirus disease 2019 (COVID-19) instances in Turkey. As a result, this analysis reveals that the geographic distribution of COVID-19 instances is associated with disparities in education, socioeconomic status, and population within individual provinces. By utilizing a composite indicator of development level based on provinces, we employ multivariate local Geary and multivariate local neighbor match tests to demonstrate the association between COVID-19 instances and the demographic and socioeconomic similarities or contrasts in provinces of Turkey. In addition, we provide an extremely randomized tree regression model to show how demographic and socioeconomic disparities affect COVID-19 instances. According to this model, the average household size, the proportion of the working-age population to the nonworking-age population, and the GDP per capita are the most important variables. The study’s main finding is that the important variables that were attained were also used to create an index for the degree of development in the Turkish regions. In other words, the same variables correlate with the degree of provincial development and the distribution of COVID-19 cases.

Explainable Artificial Intelligence Methods for Analysis of Factors Influencing Covid-19 Cases in Türki̇ye

“…Interestingly, the GDP is more prominent in the west. Provinces in the western region are at the forefront of the development index (Icoz and Ozcetin, 2023). In percentage terms, GDP is the most common of the primary factors, accounting for 31%.…”

“…The data used by Icoz and Ozcetin (2023) Breakdown plots are among the suggested tools for instance-level exploration. These plots decompose the prediction of an instance in terms of a variable to address which explanatory variable contributes more than others.…”

“…The importance ranks of the locally analyzed factors were compared with the factor importance ranks in the model. Icoz and Ozcetin (2023) reported that the number of COVID-19 cases and the development levels of provinces in Türkiye include multiple spatial autocorrelations; as a result, spatial clusters were formed. The authors also determined the importance of the factors affecting COVID-19 cases with the extra trees method, which is a modi ed version of the random trees model.…”

“…In this study, the random trees method was used for ease of application, and while examining the importance factors on a local province basis, the provinces in the clusters obtained by Icoz and Ozcetin 2023 were examined. High Moran's I values of the dependent variable and the most important variables of study (Icoz and Ozcetin 2023) according to the random forest model suggested the use of spatial models for better model t than classical machine learning models.…”