POPULATION FORECAST WITH THE DATA PROCESSING GROUP METHOD (GMDH) type NEURAL NETWORK FOR EUROPEAN UNION COUNTRIES

Abstract

Population is a critically important factor in a country's planning, policy making, and setting its social and economic goals. Population estimation and planning in advance are of great importance for policy makers, since the natural resources, which are the production areas where people can meet their basic needs, are limited and they need to protect the areas they live in in order to continue their lives. In this study, the Group Method of Data Handling (GMDH) type Neural Network (NN) approach was used for the annual population estimation of 27 European Union (EU) countries (Germany, Austria, Belgium, Bulgaria, Czech Republic, Denmark, Estonia, Finland, France, Cyprus, Croatia, Netherlands, Ireland, Spain, Sweden, Italy, Latvia, Lithuania, Luxembourg, Hungary, Malta, Poland, Portugal, Romania, Slovenia, Slovak Republic, Greece). The data set was obtained from the World Data Bank and analyzed using data from the years 1960 - 2020. The test performances obtained are generally below 10% of the Root Mean Square Percentage Error (RMSPE). The coefficient of determination (R^2) is above 0.90 and generally around 0.99. In addition, the mean absolute percentage error (MAPE) value is below 10%. According to these values, it is concluded that the model predicts extremely accurately. In addition, the analysis was compared with the 2021 - 2032 forecast values in the World Bank Database. According to the findings and comparison results, it has been concluded that the GMDH type Neural Network is a very good approach for the annual population estimation of 27 EU countries, it has almost exactly the same results with the real values in the past years, therefore it is consistent and successful in its predictions for the future years.