Artificial Neural Networks as a Method for Forecasting Migration Balance (A Case Study of the City of Lublin in Poland)

Internal migration regulates both the size and structure of human resources and affects the labor market at different spatial scales. It therefore has not only a demographic dimension, but also a spatial one, which is why it can significantly affect development on both a local and regional scale. The main objective of this study was to examine the usefulness of artificial neural networks (ANN) for predicting the internal migration balance for the city of Lublin in Poland. Another objective was to develop an experimental neural network model for forecasting the internal migration balance for the city of Lublin (for one year ahead) based on selected economic and social factors. The study area included the city of Lublin and 14 municipalities located in the vicinity of the city and functionally connected to it (they form the Lublin Functional Area), i.e., a total of 15 spatial units. Data for the analysis covered the years 2005–2022 and were obtained from the Local Data Bank (BDL) of the Central Statistical Office (GUS). The number of input variables for the ANN model was reduced using principal component analysis (PCA), allowing for the inclusion of the most relevant demographic and economic features. These components can thus be considered reliable predictors of the migration balance for the city of Lublin. This suggests that artificial neural networks may be an effective tool in supporting decision-making processes for forecasting the migration balance of this city.