Machine learning techniques often get the bad reputation of being black box methods. This session busts that myth by demonstrating how interpretability tools can give more confidence in a machine learning model and also helps to improve the insights it generates. This talk will cover best practices for using techniques such as feature importance, partial dependence, and explanation approaches. Along the way, we will consider different issues that may affect model interpretation and performance.
Understanding Your Machine Learning Model: Black Box No More
Dr. Gourab De is the VP of Data Science Practice team at DataRobot, where his primary focus is helping customers improve their ability to make and implement predictions. Dr. De's previous expertise is in the application of analytics in the pharmaceuticals, providers and payer space. His professional experience spans a wide variety of topics including genetics, bioinformatics, clinical trials, real-world effectiveness analysis with claims, PROs and charts, population health management, behavioral health, and predictive modeling. Dr. De graduated with honors from the Indian Statistical Institution in Kolkata, India, and received a Ph.D. in Biostatistics from Harvard University.