内容摘要：Machine learning stands at the frontier of today’s technological progress, and it influences the way we conduct economic research. Given the wide availability of statistical software, economic applications of machine learning methods are burgeoning. However, there remains a wide gap to fill until machine learning becomes mainstream and can be routinely employed in empirical research. The theory of machine learning is mostly established for generic statistical models, but not tailored for context of economic interest. In this talk, I will review my works with collaborators in bridging machine learning and econometrics. They are either innovative machine learning algorithms that shed light on empirical economic questions, or studies of existing machine learning methods’ properties in economic settings, in particular nonstationary time series and panel data. We work under standard econometric theoretical frameworks and make progress in asymptotic guarantee. We also develop open-source software to engage users.