Essays on applied machine learning

In this Dissertation, we deal with a series of applications of machinelearning in the fields of social and health sciences. We introduce a setof novelties in the traditional usage of machine learning algorithms forpredictive and causal inference tasks. In Part 1, we explore the field ofmachine learning for causal inference and we introduce two innovativetechniques that combine state-of-the-art machine learning algorithmswith causal inference methodologies. In the first Chapter, we introduce anovel Bayesian tree-based methodology to draw causal inference on heterogeneouseffects in quasi-experimental scenarios. In the second Chapter,we account for possible drawbacks of tree-based methodologies byproposing a composite algorithm with a high level of interpretability andprecision. In Part 2, we introduce applications of machine learning predictivepower to forecast students’ financial literacy scores and firm’s financialdistress. In the third Chapter, we innovate the applied machinelearning literature by proposing a novel sensitivity analysis for predictions.Finally, in the fourth Chapter, we show how economic intuitioncan boost the performance of machine learning algorithms. The Dissertationcontributes to the literatures on causal and predictive machine learningmainly by: (i) extending the current framework to novel scenariosand applications (Chapter 1 - Chapter 3); (ii) introducing interpretabilityin the learning models (Chapter 1 - Chapter 2); (iii) developing a novelmethodology to assess the robustness of predictions (Chapter 3); (iv) informingthe choice of the technique used by specific economic knowledgeon the field of investigation (Chapter 4). In the applied Sections ofeach Chapter, we answer policy relevant questions that pave the way tothe usage of machine learning for targeted interventions.