Machine Learning and Regularization Technique to Determine Foreign Direct Investment in Hungarian Counties

Recent studies show regional factors play an important role to attract Foreign Direct Investment (FDI) and, the performance of these factors varies within the country. Therefore, it is important to develop a measurement system to analyse the insight of these FDI factors. In this study, we used the regularization method with machine learning to get insight into the FDI determinants at the regional level. We used 18-years post-socialist period data at the county level from Hungary and applied a machine learning algorithm on different methods of regressions such as a linear, ridge, lasso, and elastic net. We analyse the relation of two dependent variables, the total amount of FDI inflow in a county and disparity of FDI inflow in companies within the county, and used urbanization, GDP per capita, labour productivity, market share of the companies, agglomeration of industries, and growth rate of the companies as predictors. Our results show that the elastic net is the best method to determine the predictive performance of FDI at the regional level.