In three papers, this dissertation (develops and) demonstrates the effective use of various machine learning (ML) tools in labor economics. The tools target different empirical purposes in the analysis of female employment. The first paper deals with data-driven classification in the analysis of maternal employment. The paper focuses on detecting latent group structures in the effect of motherhood on employment and examines how the introduction of a generous parental benefit reform impacts the different cluster groups. The second paper turns to the prediction aspect of ML in labor economics and analyzes in a data-driven way how far childbirth can be predicted from a rich set of predictor variables derived from female employment and wage histories. The third paper introduces ML tools for controlled variable selection to economists. More specifically, the paper extends a recently proposed approach for data-driven variable selection in high-dimensional linear models to the non-linear case and exemplifies its usefulness with an application towards the labor market.