| Abstract [eng] |
This thesis investigates the application of Bayesian hierarchical models and machine learning methods, such as Bayesian Additive Regression Trees (BART) and Bayesian Neural Networks (BNNs), to the modeling of distribution of firm size by employees, size classes, industries and countries. By applying hierarchical Bayesian model this analysis uses granular socio-economic dataset of indicators number of employees (EMP), turnover (TRN), and enterprise count (ENT) as these datasets exhibit complex hierarchical dependencies, missing data, and diverse scales across groups of industries and size classes. This thesis investigates the group effects on the hierarchical structure and models capability to capture the structure and estimate distribution of number of employees by different size classes. Traditional linear models are not suitable for such complexity and structure of the data motivating the use of hierarchical Bayesian model and machine learning models that are capable of incorporating prior knowledge and investigate complex data structures. |
