Abstract: 

Feature engineering is a costly process requiring a strong understanding of both the business context and data. Recently, various technologies including automated feature engineering have been developed to simplify and streamline this process. However feature engineering process as dictated by its nature remains highly exploratory and iterative. From this perspective, data scientists’ manual iteration utilizing automated feature engineering technology is presumably the most viable approach. In this presentation, we will discuss a set of best practices that combine automated feature engineering with data scientist’s manual process to iterate and monitor experiments, broadening scope of feature discovery and deepening feature understanding.

Bio: 

Yusuke Muraoka is a Principal Data Scientist at data science automation company, dotData, and has been working on enabling users to complete data science projects with bigger impact and greater efficiency, whilst improving Auto FE, Auto ML products.