摘要：
Supervised machine learning models are designed to make the most accurate predictions possible and sacrifice interpretability for predictive power. In reality, there exists an empirical trade-off such that more accurate models tend to be more complex and hence difficult to interpret. Though machine learning methods have increasingly gained popularity in recent economic, financial, and management studies, the lack of interpretability hinders its further application, especially in studies when the goal is uncover the underlying relationship between key variables instead of simply prediction. The good news is that the recent development in interpretable machine learning methods makes it possible to interpret even the complex black-box models. In this seminar, I will introduce some basic concepts of interpretable machine learning methods such as feature importance, partial dependence plots (PDP), Accumulated Local Effects (ALE), Local interpretable model-agnostic explanations (LIME), and SHAP (SHapley Additive exPlanations) values.
主讲人简介：
姚力
国际商学院助理教授，美国科罗拉多大学博士，浙江工商大学国际商学院助理教授。主要研究领域为技术与创新管理、区域与城市经济学，在Habitat International，Technological Forecasting and Social Change等国际期刊发表多篇论文。